JP4658852B2 - Adaptive block length encoding apparatus, method thereof, program and recording medium - Google Patents

Description

  The present invention relates to a time-series signal compression coding technique, and more particularly to a technique for performing coding by hierarchically dividing a frame into a plurality of blocks.

It is known that compression efficiency can be improved by adaptively selecting an analysis unit length in frame unit encoding. As one of the methods, a method is known in which the number of samples for each frame is fixed, and each frame is hierarchically divided into a plurality of blocks for encoding (see Non-Patent Document 1, for example). In the following, a conventional method for encoding by dividing each frame into a plurality of blocks hierarchically will be described first.
FIG. 16 is a block diagram illustrating the configuration of a conventional encoding apparatus that performs such an encoding process. FIG. 17 is a conceptual diagram illustrating the configuration of the frame 100 in which this encoding process is performed. A short time interval consisting of discrete times (sample points) each corresponding to a plurality of samples (usually several hundred to several thousand samples) is called a frame, but here a short time interval consisting of discrete times corresponding to all samples. Is assumed to be a frame, and the number of sample points in the frame is assumed to be about 1024 to 32768 points. In the following, the number of sample points in a frame is expressed as n. In the example of FIG. 17, one frame is divided into four layers (0th layer 110 to 3rd layer 140). In this example, in the 0th layer 110, the frame 100 matches the block B (0, 1). The first hierarchy 120 is composed of blocks B (1, 1) and B (1, 2) obtained by dividing the block B (0, 1) of the 0th hierarchy 110 into two. Further, the second hierarchy 130 divides the block B (1, 1) of the first hierarchy 120 into two blocks B (2, 1) and B (2, 2) and the block B (1, 2) into two. Block B (2,3), B (2,4). The third hierarchy 140 is divided into blocks B (3, 1) and B (3, 2) obtained by dividing the block B (2, 1) of the second hierarchy 130 into two and blocks B (2, 2) into two. Block B (3,3), B (3,4), block B (3,5), B (3,6) obtained by dividing block B (2,3) into two, and block B (2,4) ) Is divided into two blocks B (3, 7) and B (3, 8). In FIG. 17, 1, n ₁ , n ₂ , n ₃ , n ₄ ,..., N indicate numbers corresponding to the sample points in the frame 100. For example, the block B (2, 2) is a block configured by n ₂ + 1st to n _4th samples. Also, the configuration in FIG. 16 and the division method into blocks in FIG. 17 are examples, and the number of layers, the number of block divisions, and the like are not limited thereto.

In this conventional method, prediction analysis, prediction coefficient coding, prediction residual signal calculation by a prediction filter and coding (entropy coding, etc.) are independently executed for each block of each layer, and a residual code for each block is executed. Is calculated. Then, the total code amount of the residual code corresponding to the plurality of blocks in the lower layer is compared with the code amount of the residual code corresponding to the block in the upper layer composed of the plurality of blocks in the lower layer. In the frame 100, a combination of blocks having the smallest code amount is selected.
Specifically, referring to the example of FIG. 17, first, the blocks B (0, 1), B (1, 1), B (1, 2), B (2, 1) to B (2, 4), B For each (3, 1) to B (3, 8), prediction analysis, prediction coefficient coding, prediction residual signal calculation using a prediction filter, and coding thereof are performed independently to calculate a residual code for each block. Then, the code amount thereof is calculated.

  Next, the total value of the residual code amounts of the blocks B (3, 1) and B (3, 2) of the third layer 140, and the block B (2, The code amount of the residual code of 1) is compared, and the one with the smaller code amount is selected. In the example of this figure, the total code amount of the residual codes of the blocks B (3,1) and B (3,2) is larger than the code amount of the residual codes of the block B (2,1). The block B (2, 1) of the second hierarchy 130 is selected. In this figure, the selected block is shaded. Similarly, the total value of the residual code amounts of the blocks B (3, 3) and B (3, 4) of the third layer 140 and the block B (2, 2) of the second layer 130 which is an upper layer thereof. The code amount of the residual code in 2) is compared, and the one with the smaller code amount is selected. In the example of this figure, blocks B (3, 3) and B (3, 4) in the third hierarchy 140 are selected. Further, the same processing is performed for the other blocks B (2,3), B (2,4), B (3,5) to B (3,8) in the second and third hierarchies, and block selection is performed. .

Next, the sum of the code amounts of the residual codes of the selected block B (2, 1) of the second layer 130 and B (3, 3), B (3,4) of the third layer 140, and their The code amount of the residual code of the block B (1, 1) of the first layer 120 as the upper layer is compared, and the one with the smaller code amount is selected. In the example of this figure, blocks B (2,1), B (3,3), B (3,4) are selected. Block B (1,2) and its lower blocks (blocks selected from blocks B (2,3), B (2,4), B (3,5) to B (3,8)) The same block selection is performed for. In the example of this figure, the block B (1, 2) of the first hierarchy 120 is selected. Next, the sum of the code amounts of the residual codes of the blocks B (2,1), B (3,3), B (3,4), B (1,2) selected so far, and their higher order The code amount of the residual code of block B (0, 1) of the 0th layer 110 which is the layer is compared, and the one with the smaller code amount is selected. In the example of this figure, blocks B (2,1), B (3,3), B (3,4), B (1,2) are selected. Then, a prediction coefficient code, a residual code, a block corresponding to the blocks B (2,1), B (3,3), B (3,4), B (1,2) selected in this way Selection information indicating B (2,1), B (3,3), B (3,4), B (1,2) is output as a code string of the frame 100.
ISO / IEC 14496-3: 2005 / AMD2 (commonly known as MPEG-4 ALS)

As described above, in the conventional method, residual codes are obtained independently for all blocks, and their code amounts are compared to determine a hierarchical block division method. In this case, since it is necessary to calculate the code amount of the residual code for all blocks in a plurality of layers, there is a problem that the amount of calculation for encoding is very large.
The present invention has been made in view of such a point, and in adaptive block length coding that performs coding by hierarchically dividing a frame into a plurality of blocks, a technique for reducing the amount of calculation for coding The purpose is to provide.

  In the present invention, first, a plurality of hierarchies composed of one or a plurality of blocks are generated from a frame that is a time interval of an input signal (block division process), and each of the blocks obtained thereby is corresponded. The residual code amount estimation value to be calculated is calculated (residual code amount estimation process). Next, when a certain section in the frame is composed of different blocks in two or more hierarchies, the comparison result of the total value of the residual code amount estimation values of the respective hierarchies corresponding to the section is output. (Comparison process). Then, using the comparison result, select a combination of blocks constituting the frame, and the sum of the residual code amount estimation values corresponding to each block is minimized (selection process), A prediction residual signal obtained by predicting an input signal corresponding to each selected block with the first prediction order is encoded (residual encoding process). Note that the “residual code amount estimated value” in the present invention means “a value that is monotonically increasing with respect to the absolute value of the amplitude of the prediction residual signal”.

  Here, the residual code amount estimation value defined as described above has a monotonically increasing relationship with the code amount of the residual code of the corresponding prediction residual signal. Therefore, by comparing the residual code amount estimation values, it is possible to select a combination of blocks that are estimated to have a small residual code amount. Further, the amount of calculation for calculating such a residual code amount estimation value is significantly smaller than the amount of calculation for encoding the prediction residual signal. Therefore, compared with the conventional example in which residual codes of all blocks are calculated and a combination of blocks that minimizes the total code amount in the frame is selected, the processing amount for encoding can be greatly reduced.

  Preferably, in the present invention, the residual code amount estimation step is a step of calculating the residual code amount estimation value by prediction of a second prediction order that is lower than the first prediction order. In this case, since processing is performed with a low order (second prediction order) until the residual code amount estimation value is calculated, it is possible to keep the amount of computation for blocks that are not finally encoded low. it can. On the other hand, since a high order (first prediction order) is used at the stage of generating a residual code after the block to be encoded is determined, a highly accurate residual code can be generated. That is, it is possible to generate a code string with high accuracy while suppressing unnecessary calculations.

  In the present invention, it is preferable that the residual code amount estimation value is an absolute value of an amplitude of a prediction residual signal obtained by predicting an input signal of each block with the first or second prediction order, or This is the total value of the energy of the prediction residual signal in each block. “Energy” means the square value of the amplitude of a signal.

  The calculation amount of such a residual code amount estimation value is significantly smaller than the calculation amount for encoding the prediction residual signal. Also, the magnitude relationship between the residual code amount estimation values is substantially the same as the corresponding code size of the residual code. When it is estimated that the amplitude of the prediction residual signal is large, it is desirable that a value obtained by totaling the absolute values of the amplitudes of the prediction residual signal in the block is the residual code amount estimation value. This is because even if the amplitude increases, the relationship between the amplitude of the prediction residual signal and its absolute value approximates relatively well to the relationship between the amplitude of the prediction residual signal and the code amount of the residual code. On the other hand, from the viewpoint of processing efficiency, it is desirable that the value obtained by summing the energy of the prediction residual signal in the block is the residual code amount estimation value. This is because energy is a square value of the amplitude of the signal, and the square value is suitable for processing based on the least square method that is frequently used for analysis processing.

In the present invention, it is preferable that the residual code amount estimation value is a value calculated using the input signal of each block and the prediction coefficient of the first or second prediction order. In this case, it is possible to generate residual code amount estimation values and compare them without calculating a prediction residual signal. As a result, it is not necessary to calculate a prediction residual signal for a block for which no code is finally output.
Furthermore, a short-term prediction residual signal obtained by short-term prediction of the input signal corresponding to each block with the first prediction order, and a long-term prediction gain obtained by long-term prediction of the short-term prediction residual signal, May be used to calculate the residual code amount estimation value. As a result, a highly accurate residual code amount estimation value can be calculated.

  Preferably, in the present invention, when a certain section in a frame is composed of different blocks in two or more hierarchies, the difference between the sums of residual code amount estimation values of the respective hierarchies corresponding to the section is calculated. It is determined whether or not the threshold value is less than or less than a predetermined threshold (determination process). Then, when it is determined that the difference between the sums of the residual code amount estimation values of the respective layers corresponding to the section is equal to or less than a predetermined threshold, the block corresponding to the section The residual code amount estimated value by the prediction is calculated, and the comparison process is performed using the residual code amount estimated value. Note that “higher-order prediction” means prediction at an order higher than the first prediction order.

  When the prediction order is increased, the estimation accuracy of the residual code amount estimation value is improved, but the calculation amount is increased. Conversely, when the prediction order is lowered, the estimation accuracy of the residual code amount estimation value decreases, but the calculation amount decreases. By adopting a configuration in which the prediction order is increased only when the difference between the residual code amount estimation values is less than or less than a predetermined threshold, the code amount is accurately minimized without unnecessarily increasing the amount of computation. can do.

  Preferably, in the present invention, when a certain section in a frame is configured by different blocks in two or more layers, the difference between the total values of the residual code amount estimation values of the respective layers corresponding to the section. Is determined to be less than or less than a predetermined threshold (determination process). Then, when it is determined that the difference between the total values of the residual code amount estimation values of the respective layers corresponding to the section is equal to or less than a predetermined threshold value, the block corresponding to the section is respectively Each prediction residual signal obtained by predicting the corresponding input signal with the first prediction order is encoded to obtain a residual code (partial residual encoding process). In addition, for a section in which a residual code is generated by the section residual coding process, a comparison result of the total value of the code amount of the residual code of each layer corresponding to the section is output (code amount comparison). process). Then, for the section in which the residual code is generated by the section residual signal coding process, the above selection is performed using the code amount of the residual code instead of the residual code amount estimated value, and the selected block Among them, a prediction residual signal is encoded for a block for which a residual code is not obtained in the section residual signal encoding process.

  When the difference between the residual code amount estimation values is small, the magnitude relationship between the residual code amount estimation values and the magnitude relationship between the code amounts of the residual codes may be reversed due to the influence of errors or the like. Only when the difference between the residual code amount estimation values is small, the code amount of the actual residual code can be compared, thereby minimizing the code amount accurately while keeping the amount of computation as low as possible. it can.

  In the present invention, rather than comparing the code amounts obtained by actually encoding the prediction residual signals, the residual code amount estimation values are compared, and the total code amount in the frame is estimated to be the minimum. The combination of blocks to be selected is selected, and the prediction residual signal of each selected block is encoded. Thereby, in the adaptive block length coding in which the frame is hierarchically divided into a plurality of blocks and the coding is performed, the amount of calculation for the coding can be reduced.

The best mode for carrying out the present invention will be described below with reference to the drawings.
In the present invention, a plurality of hierarchies composed of one or a plurality of blocks are generated from a frame that is a time interval of an input signal, and residual code amount estimates corresponding to the respective blocks obtained thereby are obtained. calculate. Next, when a certain section in the frame is composed of different blocks in two or more hierarchies, the comparison result of the total value of the residual code amount estimation values of the respective hierarchies corresponding to the section is output. . Then, using the comparison result, a combination of blocks constituting the frame, which selects a residual code amount estimation value corresponding to each block that has the smallest sum in the frame, is selected, and the selected block The prediction residual signals obtained by predicting the input signals respectively corresponding to the first prediction order are encoded. As a result, the number of blocks that are actually encoded can be reduced, and the amount of calculation for encoding can be reduced as a whole.

  In applying the present invention, there is no particular limitation on the number of layers and the divided block length when the frame is divided into blocks, and it is not necessary that the block lengths of all the blocks are the same in one layer. For example, the present invention may be applied when a frame is divided into a plurality of blocks as shown in FIG. 17, or the present invention may be applied when a frame is divided into a plurality of blocks as shown in FIG. .

  In the dividing method of FIG. 18A, the number of blocks in the frame is the same in the first layer and the second layer, and the dividing position in each frame is different. In this case, the residual in the frame is obtained from “B (0,1)”, “B (1,1) + B (1,2)”, and “B (2,1) + B (2,2)”. The one with the smallest code amount estimation value is selected. In addition, the division method of FIG. 18B divides the first layer and the second layer as in the division method of FIG. 14A, but does not provide the 0th layer that does not divide the frame. In this case, from “B (1,1) + B (1,2)” and “B (2,1) + B (2,2)”, the sum of the residual code amount estimation values in the frame is minimum. Select what will be. In addition, in the division method of FIG. 18 (c), the division positions in the frame match only in one place between the first hierarchy and the second hierarchy, and the 0th hierarchy, the first hierarchy, and the second hierarchy match each other in the frame. The division position does not match. In this case, first, from “B (1,1) + B (1,2)” and “B (2,1) + B (2,2) + B (2,3)”, the residual code in the frame is calculated. The one with the smallest sum of quantity estimates is selected (selection 1). Also, from “B (1, 3)” and “B (2, 4) + B (2, 5)”, the one with the smallest sum of the residual code amount estimation values in the frame is selected (selected) 2). Then, the “block selected by selection 1 + the block selected by selection 2” and “B (0,1) + B (0,2)” have the smallest residual code amount estimated value. Select (selection 3).

  That is, the present invention can be applied if a part or all of the sections in a frame are configured with different blocks in a plurality of layers. However, an example in which the present invention is applied to the block division of FIG. 17 will be described below as an example. That is, first, in the frame 100, the block length of the highest layer (0th layer) matches the frame length, and the length of half the block length of the d (d = {0,1,2}) layer is The block of the (d + 1) -th layer constituting the d-th layer block is divided into four-layer blocks, and a residual code amount estimation value corresponding to each block is calculated. Then, the residual code amount estimated value corresponding to the upper layer block is compared with the sum of the residual code amount estimated values respectively corresponding to the plurality of blocks of the lower layer constituting the block, A comparison result is used to select a combination of blocks constituting a frame, which has a minimum sum of residual code amount estimation values corresponding to each block in the frame. Then, the prediction residual signal obtained by predicting the input signal corresponding to each selected block with the first prediction order is encoded.

  First, Example 1 of the present invention will be described. In the first embodiment, a prediction residual signal is calculated for each block of each layer, and a residual code amount estimation value of each block is calculated using them. Then, the residual code amount estimated value corresponding to the block in the upper layer is compared with the total value of the estimated residual code amount corresponding to the plurality of blocks in the lower layer constituting the block, and the frame is compared. Are selected from the combinations of blocks that form the sum of the residual code amount estimation values corresponding to the respective blocks, and the predicted residual signals corresponding to the selected blocks are respectively selected. Encode.

<Configuration>
FIG. 1 is a block diagram illustrating a configuration of an adaptive block length encoding apparatus 10 according to the first embodiment.
As illustrated in FIG. 1, the adaptive block length encoding apparatus 10 according to the first embodiment includes a 0th layer prediction analysis unit 10a, a 0th layer prediction coefficient encoding unit 10b, and a 0th layer that process a signal of a 0th layer block. Hierarchy prediction filter 10c and zeroth layer residual code amount estimation unit 10d, first layer prediction analysis unit 11a, first layer prediction coefficient coding unit 11b, and first layer prediction filter for processing the signals of the first layer block 11c and the first layer residual code amount estimation unit 11d, the second layer prediction analysis unit 12a, the second layer prediction coefficient coding unit 12b, the second layer prediction filter 12c, and the second layer prediction signal processing unit 12a that process the signal of the second layer block. A second layer residual code amount estimation unit 12d, a third layer prediction analysis unit 13a, a third layer prediction coefficient coding unit 13b, a third layer prediction filter 13c, and a third layer remaining Difference sign An estimation unit 13d, a division determination unit 14, a residual coding unit 15, a control unit 16, and a memory 17, a block division portion 18. Moreover, the division | segmentation determination part 14 has the comparison part 14a, the selection part 14b, and the calculating part 14c.

  The adaptive block length encoding apparatus 10 according to the first embodiment is, for example, a known computer including a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), an auxiliary storage device, and the like. A predetermined program is read and the CPU executes this program (the same applies to the following embodiments).

FIG. 2 is a conceptual diagram illustrating the data configuration of the division table 17 a stored in the memory 17. The division table 17a includes information (identifier etc.) specifying each block B (d, u) obtained by block division processing described later, and a sample (input signal) x (d) corresponding to each block B (d, u). , U, j _{d, u} ) and block position information A (d, u) for specifying the hierarchy of each block B (d, u) and the position in the frame. Note that d is a value indicating the hierarchy of the block, and the lower the value of d, the higher the hierarchy (see FIG. 17). U indicates the number of the block in the hierarchy. J _{d, u} indicates the discrete time (sample point) of each sample belonging to each block B (d, u). For example, in the case of block B (0,1) in FIG. 17, j _0,1 = 1, ..., n, and the case of the block _{B (2,2), j 2,2 =} n 2 +1, ..., a n _4.

FIG. 3 is a conceptual diagram illustrating the data configuration of the block table 17 b stored in the memory 17. The block table 17b includes a prediction residual code Cp (d, u), a prediction residual signal y (d, u, jd _{, u} ), and a residual code amount estimated value calculated as described later. It is a table in which PC (d, u) is associated with information specifying each block B (d, u).

<Processing>
Hereinafter, the adaptive block length encoding method according to the first embodiment will be described. Note that the adaptive block length encoding apparatus 10 according to the first embodiment executes each process under the control of the control unit 16. Data in each process of the adaptive block length coding apparatus 10 is read and written to the memory 17 one by one, but in principle, the description thereof is omitted below. Furthermore, the order of processing to be described is merely an example, and it is needless to say that changes can be made as appropriate without departing from the spirit of the present embodiment. Furthermore, at least a part of the processing may be executed in parallel (the same applies to the following embodiments). The input signal to the adaptive block length coding apparatus 10 is a discrete signal sampled at a predetermined sampling frequency. In addition, each process is executed for each frame that is a short time section composed of discrete times (sample points) respectively corresponding to a plurality of samples. Here, a short time interval (for example, having a sample point of about 1024 to 32768) corresponding to all samples may be used as a frame, and a part of the short time interval (for example, the number of sample points is Some hundreds to thousands of points) may be used as a frame. In this embodiment, n is the number of sample points in one frame. In addition, when a part of a short time section of all sample points is a frame, each process is repeated for each frame. However, only one frame process will be described below (the following examples are also described). The same).

[Block division processing]
First, an input signal for one frame including n samples is input to the block dividing unit 18. The block dividing unit 18 hierarchically divides this frame into a plurality of blocks (see FIG. 17), and samples x (d, u, j _{d, u} ) of each divided block B (d, u), The block position information A (d, u) of each block is stored in the division table 17a of the memory 17 in association with information specifying the block B (d, u). Here, “block division” means a process of dividing a sample belonging to a frame into a plurality of blocks. Specifically, for example, a process for determining a sample point corresponding to each block, a start of each block, or the like. This means processing for determining sample points and end sample points (block delimiters). Note that the frame block division method (block length, number of hierarchies, etc.) may be set in the block division unit 18 in advance, or may be determined each time according to an input signal or the like.

[Processing of Level 0]
Next, the 0-layer prediction analysis unit 10a refers to, for example, block identification information of the partition table 17a of the memory 17, and samples x (0, 1,) of the block B (0, 1) of the 0th layer from the partition table 17a. j _0,1 ) (j _0,1 = 1, ..., n) is read. Then, the 0-layer prediction analysis unit 10a performs a prediction analysis of the read sample x (0,1, j _0,1 ) and performs a prediction coefficient α (0,1, i) (i = 1,..., P ) Is calculated and output. Note that p is the predicted order (corresponding to “first predicted order”). p may be set in advance or may be calculated each time according to the input signal. Further, linear prediction analysis is generally used for prediction analysis, but any prediction analysis method may be applied. For details of the prediction analysis, for example, “Takehiro Moriya,“ Speech Coding ”, The Institute of Electronics, Information and Communication Engineers, ISBN 4-88552-156-4,“ Author Sadahiro Furui, “Acoustic / Speech Engineering” ", Modern Science Co., Ltd., ISBN 4-7649-0196-X", etc., and the description thereof is omitted.

  Next, the 0-layer prediction coefficient encoding unit 10b quantizes the input prediction coefficient α (0,1, i) and the quantization prediction coefficient α ^ (0,1, i). The prediction coefficient code Cp (0,1) of ^ (0,1, i) is obtained and these are output. The prediction coefficient code Cp (0, 1) output here is stored in the block table 17b of the memory 17 in association with information specifying the corresponding block B (0, 1) (see FIG. 3).

Next, the 0th layer prediction filter 10c uses the input quantized prediction coefficient α ^ (0,1, i) and uses the sample x (0,1, j _0,1 ) of the block B ( _0,1 ). To obtain a prediction residual signal y (0,1, j _0,1 ) and output it. The prediction residual signal y (0,1, j _0,1 ) output in this way is stored in the block table 17b of the memory 17 in association with the information specifying the corresponding block B (0,1). (See FIG. 3). Note that filtering here means processing for obtaining a difference between a prediction model (filter) into which a quantized prediction coefficient is substituted and an input signal.

Next, the 0th layer residual code amount estimation unit 10d uses the prediction residual signal y (0,1, j _0,1 ) read from the block table 17b of the memory 17 to generate the block B (0,1). The corresponding residual code amount estimation value PC (0, 1) is calculated and output. The output residual code amount estimation value PC (0, 1) is stored in the block table 17b of the memory 17 in association with the corresponding block B (0, 1) (see FIG. 3). In the present invention, the difference between the residual code amount estimated value corresponding to a block of a certain layer and the total value of the residual code amount estimated values respectively corresponding to a plurality of blocks of other layers constituting the block is determined. In comparison, a combination of blocks constituting a frame, which selects a residual code amount estimation value corresponding to each block, which is the smallest in the frame, is selected. Therefore, the residual code amount estimated value does not have to be a value that estimates the residual code itself, but is estimated to be in a monotonically increasing relationship with the code amount when the prediction residual signal is encoded. If it is. In this embodiment, the absolute value of the amplitude of the prediction residual signal or the value of the energy of the prediction residual signal (the square of the amplitude of the prediction residual signal) is summed within the block as the residual code amount estimation value. . Note that whether to use the sum of absolute values of amplitude or the sum of energy is unified between blocks (the same applies to the following embodiments).

[First level processing]
Similarly, the first layer prediction analysis unit 11a, the first layer prediction coefficient coding unit 11b, the first layer prediction filter 11c, and the first layer residual code amount estimation unit 11d are connected to the block B (1, 1) and B (1,2), independently of each other, prediction coefficient codes Cp (1,1), Cp (1,2) and prediction residual signal y (0,1, j _1,1 ), y (0,1, j _1,2 ) and residual code amount estimation values PC (1,1), PC (1,2) are calculated. Each piece of calculated information is associated with information for identifying the corresponding block B (1,1), B (1,2) and stored in the block table 17b of the memory 17 (see FIG. 3).

[Second level processing]
Similarly, for the second layer, the second layer prediction analysis unit 12a, the second layer prediction coefficient coding unit 12b, the second layer prediction filter 12c, and the second layer residual code amount estimation unit 12d are connected to the second layer block. A prediction coefficient code, a prediction residual signal, and a residual code amount estimation value are calculated independently for B (2,1) to B (2,4). Each piece of calculated information is associated with information specifying the corresponding block B (2, 1) to B (2, 4), and is stored in the block table 17b of the memory 17 (see FIG. 3).

[Third layer processing]
Similarly, for the third layer, the third layer prediction analysis unit 13a, the third layer prediction coefficient coding unit 13b, the third layer prediction filter 13c, and the third layer residual code amount estimation unit 13d are arranged in the third layer. For each block B (3, 1) to B (3, 8), a prediction coefficient code, a prediction residual signal, and a residual code amount estimation value are calculated independently. Each piece of calculated information is associated with information for specifying the corresponding block B (3, 1) to B (3, 8) and stored in the block table 17b of the memory 17 (see FIG. 3).

[Processing of division determination unit 14]
Next, the comparison unit 14a of the division determination unit 14 reads the residual code amount estimated value corresponding to each block of each layer from the block table 17b of the memory 17, and the residual code amount estimated value corresponding to the block of the upper layer. And the total value of the residual code amount estimation values respectively corresponding to a plurality of blocks in the lower hierarchy constituting the block are compared. Next, the selection unit 14b uses the comparison result to select a combination of blocks that constitute a frame and that has the smallest sum of residual code amount estimation values corresponding to each block in the frame. To do. Then, the selection unit 14b outputs a prediction coefficient code corresponding to each selected block, a prediction residual signal, and selection information indicating a combination of the selected blocks.

A specific example of this processing procedure will be described below. Note that the processing procedure shown below is merely an example, and can be appropriately changed without departing from the gist of the present invention.
FIG. 4 is a flowchart for explaining an example of processing of the division determination unit 14. Hereinafter, an example of the process of the division determination unit 14 will be described with reference to FIG. This processing example is for the case where block division is performed as shown in FIG. 17, but it is easy to extend this to other hierarchical block division.

First, the selection unit 14b of the division determination unit 14 searches the block specifying information in the division table 17a of the memory 17, and each block B (3, r) (r = 1,. , 2 ³ ) are elements of the set S. And the calculating part 14c stores the information which shows this set S in the memory 17 (step S1). Next, the computing unit 14c substitutes 2 (the number of the next highest hierarchy in the lowest hierarchy) for the variable d, and stores this variable d in the memory 17 (step S2).

  Next, the computing unit 14c substitutes 1 for the variable u and stores the variable u in the memory 17 (step S3). Next, the calculation unit 14c refers to the information indicating the set S of the memory 17 and the division table 17a, and configures the u-th block B (d, u) of the d-th layer, the plurality of blocks of the lower layer Among them, the sum of the residual code amount estimation values corresponding to those included in the set S is obtained and substituted into the variable TPC (d, u) (step S4). This variable TPC (d, u) is stored in the memory 17.

Next, the comparison unit 14a searches the block table 17b of the memory 17 using information for specifying the block B (d, u), and the residual code amount estimated value PC corresponding to the block B (d, u). (D, u) is read. Further, the comparison unit 14a reads the variable TPC (d, u) from the memory 17,
TPC (d, u)> PC (d, u)
The comparison result is output to the selection unit 14b (step S5).

  Only when the comparison result indicates that TPC (d, u)> PC (d, u), the selection unit 14b sets the set S based on the information specifying the block B (d, u). Is updated and stored in the memory 17 (step S6). Note that “update the set S with information specifying the block B (d, u)” means that the elements of the set S corresponding to the blocks in the lower hierarchy of the block B (d, u) are the blocks B (d, u). It means replacing with the element of u). In the example of FIG. 17, for example, when the set S = {B (3,1), B (3,2), B (3,3), ..., B (3,8)} In addition, the element {B (3,1), B (3,2)} of the set S corresponding to the block in the lower hierarchy constituting the block B (2,1) is used as the element of the block B (2,1). This means that the set S is updated as S = {B (2,1), B (3,3),..., B (3,8)}.

Thereafter, the memory comparison section 14a, it is determined whether the variable u of the memory 17 is ^{2 d} (step S7), and if not u is ^{2 d,} the calculation portion 14c is u + 1 as a new value of the variable u After storing in 17, the control unit 16 returns the process to step S4. On the other hand, if the variable u is a ^{2 d,} then comparing unit 14a determines whether or not the variable d in the memory 17 is 0 (step S8). If it is determined in step S8 that d is not 0, the calculation unit 14c stores d-1 as a new variable d value in the memory 17, and then the control unit 16 returns the process to step S3. On the other hand, if it is determined in step S8 that d is 0, the selection unit 14b reads the set S from the memory 17, and predicts the prediction coefficient code and the prediction residual signal associated with each block indicated by the elements of the set S. Are read from the block table 17b of the memory 17. Then, the selection unit 14b outputs the read prediction coefficient code, the prediction residual signal, and selection information indicating the elements of the set S. When the decoding apparatus does not hold a correspondence table between blocks and block identification information, the elements of the set S and the block identification information corresponding to the elements may be selected information.

[Processing of Residual Encoding Unit 15]
Each prediction residual signal output from the selection unit 14b is input to the residual encoding unit 15, and the residual encoding unit 15 outputs the prediction residual signal for each block selected by the selection unit 14b. An encoded residual code is obtained, and the residual code is output. As an encoding method, for example, entropy encoding or the like is used (the same applies to the following encoding).
Then, the adaptive block length encoding device 10 outputs the residual code output from the residual encoding unit 15, the prediction coefficient code output from the selection unit 14b, and the selection information as a code string.

<Characteristics of Example 1>
Residual codes calculated by the 0th layer residual code amount estimating unit 10d, the first layer residual code amount estimating unit 11d, the second layer residual code amount estimating unit 12d, and the third layer residual code amount estimating unit 13d. The quantity estimation value is the absolute value of the amplitude of the prediction residual signal or a value obtained by summing up the energy of the prediction residual signal in the block. The amount of calculation for calculating this value is considerably smaller than the processing amount required for encoding such as entropy encoding. Therefore, as a whole, the processing amount can be greatly reduced as compared with the conventional configuration. Further, the magnitude of the absolute value of the amplitude of the prediction residual signal or the magnitude of the sum of the energy of the prediction residual signal is substantially the same as the code amount of the actual residual code. Therefore, a combination of blocks having a large code amount is selected due to erroneous estimation, and the overall compression code amount hardly increases.

  Next, a second embodiment of the present invention will be described. In the second embodiment, a residual code amount estimation value is calculated using an input signal corresponding to each block and a prediction coefficient obtained by predictive analysis of the input signal. As a result, the quantization / encoding processing by the prediction coefficient encoding unit and the number of filtering processes by the prediction filter can be reduced by (layer number-1) layers compared to the first embodiment. The number of prediction coefficient codes and prediction residual signals to be stored can also be reduced. In the following, items common to the first embodiment will be described in a simplified manner.

<Configuration>
FIG. 5 is a block diagram illustrating a configuration of the adaptive block length encoding device 20 according to the second embodiment. In addition, about the part which is common in Example 1 already demonstrated in FIG. 5, the same code | symbol as used so far was attached | subjected.
As illustrated in FIG. 5, the adaptive block length coding apparatus 20 according to the second embodiment includes a 0th layer prediction analysis unit 10a and a 0th layer residual code amount estimation unit 20d that process a signal of a 0th layer block, The first layer prediction analysis unit 11a and the first layer residual code amount estimation unit 21d that process the signal of the first layer block, and the second layer prediction analysis unit 12a and the second layer prediction analysis unit 12a that process the signal of the second layer block A 2-layer residual code amount estimation unit 22d, a third-layer prediction analysis unit 13a and a third-layer residual code amount estimation unit 23d that process signals of blocks in the third layer, a division determination unit 24, and a prediction coefficient code And a prediction filter 28, a residual encoding unit 15, a control unit 16, and a memory 17. Moreover, the division | segmentation determination part 24 has the comparison part 24a, the selection part 24b, and the calculating part 24c.

  FIG. 6 is a conceptual diagram illustrating the data configuration of the block table 27 b of the second embodiment stored in the memory 17. The block table 27b uses the prediction coefficient α (d, u, i) and the residual code amount estimated value PC (d, u) calculated as described later as information for identifying each block B (d, u). It is the table which matched.

<Processing>
Hereinafter, an adaptive block length encoding method according to the second embodiment will be described.
[Block division processing]
First, the same block division processing as that in the first embodiment is performed, and a division table 17 a is generated and stored in the memory 17.

[Processing of Level 0]
Next, the 0-layer prediction analysis unit 10a refers to, for example, block identification information of the partition table 17a of the memory 17, and samples x (0, 1,) of the block B (0, 1) of the 0th layer from the partition table 17a. j _0,1 ) (j _0,1 = 1, ..., n) is read. Then, the 0-layer prediction analysis unit 10a performs a prediction analysis of the read sample x (0,1, j _0,1 ) and performs a prediction coefficient α (0,1, i) (i = 1,..., P ) Is calculated and output. The output prediction coefficient α (0, 1, i) is stored in the block table 27b of the memory 17 in association with information specifying the corresponding block B (0, 1) (see FIG. 6).

Next, the 0th layer residual code amount estimator 20d calculates the prediction coefficient α (0, 1, i) and the sample x (0) of the block B (0, 1) of the 0th layer read from the division table 17a. , 1, j _0,1 ), the residual code amount estimated value PC (0, 1) corresponding to the block B (0, 1) is calculated and output. The output residual code amount estimation value PC (0, 1) is stored in the block table 27b of the memory 17 in association with information specifying the corresponding block B (0, 1) (see FIG. 6).
Hereinafter, the calculation procedure of the residual code amount estimation value of the block B (0, 1) in the present embodiment will be exemplified.

[When the estimated value of the sum of the energy in the block is the residual code amount estimated value]
When the estimated value of the sum of the energy in the block is used as the residual code amount estimated value, first, the 0th layer residual code amount estimating unit 20d performs the sample x (0, 1, j) of the block B (0, 1). _{x (0,1, j 0,1) |} | energy _0,1) obtaining the sum F of the ^second block. Next, the 0th layer residual code amount estimation unit 20d converts the partial autocorrelation coefficient (PARCOR coefficient) from the prediction coefficient α (0, 1, i) to k (0, 1, i) (i∈ {1, 2, ..., p}). Note that the prediction coefficient and the partial autocorrelation coefficient are equivalent parameters, and their relationship and conversion method are also known. Next, the 0th layer residual code amount estimation unit 20d calculates a residual code amount estimated value PC (0, 1) using the following equation.

[When the estimated value of the sum of absolute values of amplitude in the block is the residual code amount estimated value]
When the estimated value of the sum of the amplitudes in the block is used as the residual code amount estimated value, first, the 0th layer residual code amount estimating unit 20d first performs the sample x (0, 1) of the block B (0, 1). , J _0,1 ) find the sum F in the block of the absolute value | x (0,1, j _0,1 ) | Next, the 0th layer residual code amount estimation unit 20d converts the partial autocorrelation coefficient (PARCOR coefficient) from the prediction coefficient α (0, 1, i) to k (0, 1, i) (i∈ {1, 2, ..., p}). Next, the 0th layer residual code amount estimation unit 20d calculates a residual code amount estimated value PC (0, 1) using Expression (1). More preferably, the 0th layer residual code amount estimation unit 20d calculates a residual code amount estimated value PC (0, 1) using the following equation. By using Expression (2), it is possible to calculate the residual code amount estimated value PC (0, 1) with higher estimation accuracy.

[Processing from the first to the third layer]
Similarly to the 0th hierarchy, the 1st to 3rd hierarchy also includes a 1st hierarchy prediction analysis unit 11a, a 1st hierarchy residual code amount estimation unit 21d, a 2nd hierarchy prediction analysis unit 12a, and a 2nd hierarchy residual code amount estimation unit. 22d, a prediction coefficient and a residual code amount estimation value for each block of each layer are calculated by the processes of the third layer prediction analysis unit 13a and the third layer residual code amount estimation unit 23d, respectively. The calculated prediction coefficient and residual code amount estimated value are each associated with information for identifying the corresponding block and stored in the block table 27b of the memory 17 (see FIG. 6).

[Processing of Division Determination Unit 24]
Next, the comparison unit 24a of the division determination unit 24 reads the residual code amount estimated value corresponding to each block of each layer from the block table 27b of the memory 17, and the residual code amount estimated value corresponding to the block of the higher layer. And the total value of the residual code amount estimation values respectively corresponding to a plurality of blocks in the lower hierarchy constituting the block are compared. Next, the selection unit 24b uses the comparison result to select a combination of blocks constituting the frame, and the sum of the residual code amount estimation values corresponding to the respective blocks is minimized. To do. The processing of the division determination unit 24 so far is performed in the same manner as the division determination unit 14 of the first embodiment, for example. Then, the selection unit 24b reads the prediction coefficient and the sample (input signal) corresponding to each selected block from the block table 27b (FIG. 6) and the division table 17a (FIG. 2) of the memory 17, respectively. And selection information indicating the combination of the blocks.

[Encoding process]
Next, the prediction coefficient encoding unit 26 selects the quantized prediction coefficient obtained by quantizing each prediction coefficient output from the selection unit 24b and the prediction coefficient code of the quantized prediction coefficient by the selection unit 24b. It calculates | requires for every block and outputs the said quantization prediction coefficient and prediction coefficient code | symbol. Next, the prediction filter 28 filters the sample (input signal) output from the selection unit 24b for each block selected by the selection unit 24b using the quantized prediction coefficient output from the prediction coefficient encoding unit 26. The prediction residual signal is obtained and output. Then, the residual encoding unit 15 generates a residual code obtained by encoding the prediction residual signal output from the prediction filter 28 for each block selected by the selection unit 24b, and outputs the residual code. .
Then, the adaptive block length encoding device 20 includes the residual code output from the residual encoding unit 15, the prediction coefficient code output from the prediction coefficient encoding unit 26, and the selection information output from the selection unit 24b. Are output as a code string.

<Features of Example 2>
In the present embodiment, the number of processes of the prediction coefficient encoding unit and the prediction filter can be reduced by (layer number−1) layers, respectively, compared to the first embodiment. And since the calculation amount of the prediction coefficient encoding unit and the prediction filter is significantly larger than the calculation amount in the 0th to 3rd layer residual code amount estimation units 20d to 23d, the calculation processing amount is smaller than that in the first embodiment. Encoding can be performed. Further, since it is not necessary to store prediction residual signals corresponding to all blocks in the block table, the capacity of the memory 17 required for the encoding process may be small.

  Next, Embodiment 3 of the present invention will be described. In the third embodiment, an estimated residual code amount is calculated using an input signal corresponding to each block and a quantized prediction coefficient of a prediction coefficient obtained by predicting and analyzing the input signal. Thereby, the number of filtering processes by the prediction filter can be reduced by (the number of hierarchies minus 1) hierarchies compared to the first embodiment, and the number of prediction residual signals to be stored in the memory 17 can also be reduced. In the following, items common to the above-described embodiments will be described in a simplified manner.

<Configuration>
FIG. 7 is a block diagram illustrating a configuration of the adaptive block length encoding device 30 according to the third embodiment. In addition, the same code | symbol as what was used until now is attached | subjected about the part which is common in the Example already demonstrated in FIG.

  As illustrated in FIG. 7, the adaptive block length coding apparatus 30 according to the third embodiment includes a 0th layer prediction analysis unit 10a, a 0th layer prediction coefficient coding unit 10b, and a 0th layer prediction coefficient coding unit 10b that process a signal of a 0th layer block. A 0-layer residual code amount estimation unit 30d, a first-layer prediction analysis unit 11a, a first-layer prediction coefficient encoding unit 11b, and a first-layer residual code amount estimation unit 31d that process signals of blocks in the first layer; The second layer prediction analysis unit 12a, the second layer prediction coefficient coding unit 12b, and the second layer residual code amount estimation unit 32d that process the signal of the second layer block, and the signal of the block of the third layer A third layer prediction analysis unit 13a, a third layer prediction coefficient coding unit 13b, a third layer residual code amount estimation unit 33d, a division determination unit 34, a prediction filter 36, and a residual coding unit 15, It has a control unit 16 and a memory 17 That. Moreover, the division | segmentation determination part 34 has the comparison part 34a, the selection part 34b, and the calculating part 34c.

  FIG. 8 is a conceptual diagram illustrating the data configuration of the block table 37 b of the third embodiment stored in the memory 17. The block table 37b includes a prediction coefficient α (d, u, i), a residual code amount estimated value PC (d, u), and a quantized prediction coefficient α ^ (d, u, i) calculated as described later. This is a table associated with information specifying each block B (d, u).

<Processing>
Hereinafter, an adaptive block length encoding method according to the third embodiment will be described.
[Block division processing]
First, the same block division processing as that in the first embodiment is performed, and a division table 17 a is generated and stored in the memory 17.

[Processing of Level 0]
Next, the 0-layer prediction analysis unit 10a refers to, for example, block identification information of the partition table 17a of the memory 17, and samples x (0, 1,) of the block B (0, 1) of the 0th layer from the partition table 17a. j _0,1 ) (j _0,1 = 1, ..., n) is read. Then, the 0-layer prediction analysis unit 10a performs a prediction analysis of the read sample x (0,1, j _0,1 ) and performs a prediction coefficient α (0,1, i) (i = 1,..., P ) Is calculated and output.

  Next, the 0-layer prediction coefficient encoding unit 10b quantizes the input prediction coefficient α (0,1, i) and the quantization prediction coefficient α ^ (0,1, i). The prediction coefficient code Cp (0,1) of ^ (0,1, i) is obtained and these are output. The prediction coefficient code Cp (0, 1) output here is stored in the block table 37b of the memory 17 in association with the information specifying the corresponding block B (0, 1) (see FIG. 8).

Next, the 0th layer residual code amount estimation unit 30d samples the quantization prediction coefficient α ^ (0,1, i) and the block B (0,1) of the 0th layer read from the division table 17a. Using x (0,1, j _0,1 ), a residual code amount estimated value PC (0,1) corresponding to block B (0,1) is calculated and output. The calculation procedure of the residual code amount estimation value is the same as that described in the second embodiment except that the prediction coefficient is replaced with the quantized prediction coefficient. The output residual code amount estimation value PC (0, 1) is associated with information specifying the corresponding block B (0, 1) and stored in the block table 37b of the memory 17 (see FIG. 8).

[Processing from the first to the third layer]
Similarly to the 0th hierarchy, the 1st to 3rd hierarchy also includes a first hierarchy prediction analysis unit 11a, a first hierarchy prediction coefficient encoding unit 11b, a first hierarchy residual code amount estimation unit 31d, and a second hierarchy prediction analysis unit. 12a, second layer prediction coefficient coding unit 12b and second layer residual code amount estimation unit 32d, third layer prediction analysis unit 13a, third layer prediction coefficient coding unit 13b, and third layer residual code amount estimation By each process by the unit 33d, a prediction coefficient code, a quantized prediction coefficient, and a residual code amount estimation value for each block of each layer are calculated. The calculated prediction coefficient code, quantized prediction coefficient, and residual code amount estimated value are each associated with information for identifying the corresponding block and stored in the block table 37b of the memory 17 (see FIG. 8).

[Processing of division determination unit 34]
Next, the comparison unit 34a of the division determination unit 34 reads the residual code amount estimated value corresponding to each block of each layer from the block table 37b of the memory 17, and the residual code amount estimated value corresponding to the block of the higher layer. And the total value of the residual code amount estimation values respectively corresponding to a plurality of blocks in the lower hierarchy constituting the block are compared. Next, the selection unit 34b uses the comparison result to select a combination of blocks that constitute a frame and that has the smallest sum of residual code amount estimates corresponding to each block in the frame. To do. The processing of the division determination unit 34 so far is performed in the same manner as the division determination unit 14 of the first embodiment, for example. Then, the selection unit 34b reads the prediction coefficient code and the quantized prediction coefficient corresponding to each selected block from the block table 37b (FIG. 8) of the memory 17. The selection unit 34b reads a sample corresponding to each selected block from the division table 17a (FIG. 2) of the memory 17. Then, the selection unit 34b outputs each read information and selection information indicating a combination of the selected blocks.

[Encoding process]
Next, the prediction filter 36 filters the sample (input signal) output from the selection unit 34b using the quantized prediction coefficient output from the selection unit 34b for each block selected above, and the prediction residual. Find the signal and output it. Then, the residual encoding unit 15 generates a residual code obtained by encoding the prediction residual signal output from the prediction filter 28 for each block selected by the selection unit 34b, and outputs the residual code. .
Then, the adaptive block length encoding device 30 outputs the residual code output from the residual encoding unit 15 and the prediction coefficient code and selection information output from the selection unit 34b as a code string.

<Features of Example 3>
In the present embodiment, the number of prediction filter processes can be reduced by (the number of hierarchies minus 1) layers as compared with the first embodiment. Since the calculation amount of the prediction filter is significantly larger than the calculation amount in the 0th to 3rd layer residual code amount estimation units 20d to 23d, encoding can be performed with a smaller calculation processing amount than in the first embodiment. Further, since it is not necessary to store a prediction residual signal having a large amount of data in the block table for all blocks, the capacity of the memory 17 required for the encoding process may be small.

  Next, a fourth embodiment of the present invention will be described. In the fourth embodiment, the residual code amount estimation values corresponding to the prediction residual signals for which the low-order (second prediction order) prediction is performed are compared, the combination of blocks to be encoded is selected, and the higher order A prediction residual signal for which prediction of (first prediction order) has been performed is encoded to generate a residual code. In the following, items common to the already described embodiments will be described in a simplified manner.

<Configuration>
FIG. 9 is a block diagram illustrating a configuration of the adaptive block length encoding device 40 according to the fourth embodiment. In addition, the same code | symbol as what was used until now is attached | subjected about the part which is common in Example 1 already demonstrated in FIG.
As illustrated in FIG. 9, the adaptive block length encoding device 40 according to the fourth embodiment includes a 0th layer lower-order prediction analysis unit 40a and a 0th layer residual code amount estimation unit that process a signal of a block of the 0th layer. 20d, a first layer lower-order prediction analysis unit 41a and a first layer residual code amount estimation unit 21d that process a signal of a first layer block, and a second layer lower-order that processes a signal of a second layer block The prediction analysis unit 42a and the second layer residual code amount estimation unit 22d, the third layer lower-order prediction analysis unit 43a and the third layer residual code amount estimation unit 23d that process the signals of the third layer block, and the division The determination unit 44, the higher-order prediction analysis unit 46, the prediction coefficient encoding unit 49, the prediction filter 48, the residual encoding unit 15, the control unit 16, and the memory 17 are included. The division determination unit 44 includes a comparison unit 44a, a selection unit 44b, and a calculation unit 44c.
FIG. 10 is a conceptual diagram illustrating the data configuration of the block table 47 b of the fourth embodiment stored in the memory 17. The block table 47b includes a low-order prediction coefficient α (d, u, i ′) (i ′ = 1,..., Q, q <p) and a residual code amount estimated value PC ( d, u) is a table in which information specifying each block B (d, u) is associated.

<Processing>
Hereinafter, an adaptive block length encoding method according to the fourth embodiment will be described.
[Block division processing]
First, the same block division processing as that in the first embodiment is performed, and a division table 17 a is generated and stored in the memory 17.
[Processing of Level 0]
Next, the 0th layer low-order prediction analysis unit 40a refers to, for example, block identification information of the partition table 17a of the memory 17, and samples x (0,1) of the block B (0, 1) of the 0th layer from the partition table 17a. 1, j _0,1 ) (j _0,1 = 1, ..., n) is read. Then, the 0-th layer low-order prediction analysis unit 40a performs prediction analysis up to the low-order (second prediction order q) of the read sample x (0,1, j _0,1 ), and performs a low-order prediction coefficient α ( 0, 1, i ′) (i ′ = 1,..., Q) are calculated and output. The output low-order prediction coefficient α (0, 1, i ′) is stored in the block table 47b of the memory 17 in association with information specifying the corresponding block B (0, 1) (see FIG. 10). .

Next, the 0th layer residual code amount estimation unit 20d samples the low-order prediction coefficient α (0, 1, i ′) and the block B (0, 1) of the 0th layer read from the division table 17a. Using x (0,1, j _0,1 ), a residual code amount estimated value PC (0,1) corresponding to block B (0,1) is calculated and output. The output residual code amount estimated value PC (0, 1) is stored in the block table 47b of the memory 17 in association with information specifying the corresponding block B (0, 1) (see FIG. 10). The calculation procedure of the residual code amount estimation value is the same as that described in the second embodiment except that the prediction coefficient is replaced with a low-order prediction coefficient.

[Processing from the first to the third layer]
Similarly to the 0th hierarchy, the 1st to 3rd hierarchy also includes a 1st hierarchy low order prediction analysis unit 41a, a 1st hierarchy residual code amount estimation unit 21d, a 2nd hierarchy low order prediction analysis unit 42a, and a 2nd hierarchy residual. By the processes of the code amount estimation unit 22d, the third layer low-order prediction analysis unit 43a, and the third layer residual code amount estimation unit 23d, the low-order prediction coefficient and the residual code amount estimation value for each block of each layer are obtained. Each is calculated. The calculated low-order prediction coefficient and residual code amount estimated value are each associated with information for specifying the corresponding block and stored in the block table 47b of the memory 17 (see FIG. 10).

[Processing of Division Determination Unit 44]
Next, the comparison unit 44a of the division determination unit 44 reads the residual code amount estimated value corresponding to each block of each layer from the block table 47b of the memory 17, and the residual code amount estimated value corresponding to the block of the upper layer. And the total value of the residual code amount estimation values respectively corresponding to a plurality of blocks in the lower hierarchy constituting the block are compared. Next, the selection unit 44b uses the comparison result to select a combination of blocks that constitute a frame and that has a minimum sum of residual code amount estimation values corresponding to each block in the frame. To do. The processing of the division determination unit 44 so far is performed in the same manner as the division determination unit 14 of the first embodiment, for example. Then, the selection unit 44b reads the low-order prediction coefficient and the sample (input signal) corresponding to each selected block from the block table 47b (FIG. 10) and the division table 17a (FIG. 2) of the memory 17, These and selection information indicating the combination of the selected blocks are output.

[Encoding process]
Next, the high-order prediction analysis unit 46 uses the low-order prediction coefficient output from the selection unit 44b and the sample (input signal), and uses the high-order prediction coefficient coefficient up to the high order (first prediction order p> q). Are obtained for each block selected by the selection unit 44b and are output. In this embodiment, since the low-order prediction coefficients have already been obtained up to the low order (second prediction order q), only the remaining prediction coefficients from the q + 1st order to the p-order need be obtained.

Next, the prediction coefficient encoding unit 49 selects the quantized prediction coefficients obtained by quantizing the high prediction coefficient coefficients output from the high-order prediction analysis unit 46 and the prediction coefficient codes thereof by the selection unit 44b. It calculates | requires for every block and outputs the said quantization prediction coefficient and prediction coefficient code | symbol. Next, the prediction filter 48 filters the sample (input signal) output from the selection unit 44b for each block selected by the selection unit 44b using the quantized prediction coefficient output from the prediction coefficient encoding unit 49. Then, a prediction residual signal is obtained and output. Then, the residual encoding unit 15 generates a residual code obtained by encoding the prediction residual signal output from the prediction filter 48 for each block selected by the selection unit 44b, and outputs the residual code. .
Then, the adaptive block length encoding device 40 includes the residual code output from the residual encoding unit 15, the prediction coefficient code output from the prediction coefficient encoding unit 49, and the selection information output from the selection unit 44b. Are output as a code string.

<Features of Example 4>
In the present embodiment, residual code amount estimation values corresponding to prediction residual signals for which low-order (second prediction order) prediction has been performed are compared, a combination of blocks to be encoded is selected, and a higher order is selected. The prediction residual signal on which the prediction of (first prediction order) has been performed is encoded to generate a residual code. Thereby, the calculation amount for selecting the combination of blocks to be encoded can be made smaller than that in the first embodiment.

  As a modification of the fourth embodiment, a configuration in which a prediction residual signal is obtained with a low-order order as in the first embodiment and a residual code amount estimation value is calculated may be employed. In the fourth embodiment, the low-order prediction coefficients calculated by the 0th to 3rd layer low-order prediction analysis units 40a to 43a are stored in the block table 47b, and the high-order prediction analysis unit 46 stores the remaining q + 1 orders. Only the prediction coefficients from the first to the pth order are obtained. However, the lower order prediction coefficients calculated by the 0th to 3rd layer lower order prediction analysis units 40a to 43a are not stored in the block table 47b, and the higher order prediction analysis unit 46 obtains prediction coefficients from the first order to the pth order. May be.

  Next, a fifth embodiment of the present invention will be described. In Example 5, after calculating a short-term coefficient code and a short-term prediction residual signal by short-term prediction analysis (for example, linear prediction analysis), long-term prediction analysis (pitch prediction analysis) of the short-term prediction residual signal is performed to perform long-term prediction. Delay amount (pitch period) and long-term prediction gain are calculated. Then, the long-term prediction residual code amount is estimated using the short-term residual code amount estimation value estimated using the short-term prediction residual signal and the long-term prediction gain, and a combination of blocks to be encoded is selected. In the following, items common to the already described embodiments will be described in a simplified manner.

<Configuration>
FIG. 11 is a block diagram illustrating a configuration of the adaptive block length encoding device 50 according to the fifth embodiment. In addition, the same code | symbol as what was used until now is attached | subjected about the part which is common in the Example already demonstrated in FIG.

  As illustrated in FIG. 11, the adaptive block length encoding device 50 according to the fifth embodiment includes a 0th layer short-term prediction analysis unit 50a and a 0th layer short-term prediction coefficient encoding unit 50b that process a signal of a 0th layer block. The 0th layer short-term prediction filter 50c, the 0th layer long-term prediction analysis unit 50d, the 0th layer residual code amount estimation unit 50e, the first layer short-term prediction analysis unit 51a that processes the signals of the first layer block, The first layer short-term prediction coefficient encoding unit 51b, the first layer short-term prediction filter 51c, the first layer long-term prediction analysis unit 51d, the first layer residual code amount estimation unit 51e, and the second layer block signal processing A second layer short-term prediction analysis unit 52a, a second layer short-term prediction coefficient coding unit 52b, a second layer short-term prediction filter 52c, a second layer long-term prediction analysis unit 52d, a second layer residual code amount estimation unit 52e, and a third Floor 3rd layer short-term prediction analysis unit 53a, 3rd layer short-term prediction coefficient coding unit 53b, 3rd layer short-term prediction filter 53c, 3rd layer long-term prediction analysis unit 53d, and 3rd layer residual code The quantity estimation unit 53e, the division determination unit 54, the long-term prediction coefficient coding unit 56, the long-term prediction filter 58, the residual coding unit 15, the control unit 16, and the memory 17 are included. The division determination unit 54 includes a comparison unit 54a, a selection unit 54b, and a calculation unit 54c.

FIG. 12 is a conceptual diagram illustrating the data configuration of the block table 57 b of the fifth embodiment stored in the memory 17. The block table 57b includes a short-term prediction coefficient code Cp (d, u), a short-term prediction residual signal y (d, u, j _{d, u} ) calculated as described later, and a long-term prediction coefficient (long-term prediction delay amount τ). (D, u), long-term prediction gain γ (d, u)) and residual code amount estimation value PC (d, u) in a table associated with information identifying each block B (d, u) is there.

<Processing>
Hereinafter, an adaptive block length encoding method according to the fifth embodiment will be described.
[Block division processing]
First, the same block division processing as that in the first embodiment is performed, and a division table 17 a is generated and stored in the memory 17.
[Processing of Level 0]
Next, the 0-layer short-term prediction analysis unit 10a refers to, for example, block identification information in the partition table 17a of the memory 17, and samples x (0, 1) of the block B (0, 1) in the 0th layer from the partition table 17a. , J _0,1 ) (j _0,1 = 1, ..., n) is read. Then, the 0-layer short-term prediction analysis unit 10a performs a short-term prediction analysis of the read sample x (0, 1, j ₀ , ₁ ) and performs a short-term prediction coefficient α (0, 1, i) (i = 1,... ., p) is calculated and output.

  Next, the 0-layer short-term prediction coefficient encoding unit 50b quantizes the input short-term prediction coefficient α (0, 1, i) and the quantized short-term prediction coefficient α ^ (0, 1, i), and the quantization The short-term prediction coefficient code Cp (0,1) of the short-term prediction coefficient α ^ (0,1, i) is obtained and output. The short-term prediction coefficient code Cp (0, 1) output here is stored in the block table 57b of the memory 17 in association with information specifying the corresponding block B (0, 1) (see FIG. 12).

Next, the 0th layer short-term prediction filter 50c uses the input quantized short-term prediction coefficient α ^ (0,1, i) and uses the sample x (0,1, j _{0, 1} ) is filtered, and the short-term prediction residual signal y (0,1, j _0,1 ) is obtained and output. The short-term prediction residual signal y (0,1, j _0,1 ) output in this way is stored in the block table 57b of the memory 17 in association with the information specifying the corresponding block B (0,1). (See FIG. 12).

Next, the 0th layer long-term prediction analysis unit 50d performs long-term prediction analysis of the short-term prediction residual signal y (0,1, j _0,1 ) output from the 0th layer short-term prediction filter 50c, and the long-term prediction delay The quantity τ (0,1) and the long-term prediction gain γ (0,1) are calculated and output. Details regarding the long-term prediction analysis are omitted because they are detailed in, for example, the half-rate speech codec (PSI-CELP) standard RCR STD-27C.
Thereafter, the 0th layer residual code amount estimation unit 50e outputs the long-term prediction gain γ (0,1) output from the 0th layer long-term prediction analysis unit 50d and the short-term prediction output from the 0th layer short-term prediction filter 50c. Using the residual signal y (0,1, j _0,1 ), the long-term residual code amount estimated value PC (0,1) corresponding to the block (0,1) is calculated and output. Hereinafter, this processing example will be specifically described by taking the block (0, 1) as an example.

[When the estimated value of the sum of the long-term prediction error energy in the block is the long-term residual code amount estimate]
When the estimated value of the sum of long-term prediction error energies in the block is used as the long-term residual code amount estimated value, first, the 0th layer long-term prediction analysis unit 50d first selects the short-term prediction residual signal y (0, 1, j _{0). ,} energy of ₁₎ | determining the ^second sum E in the block _{| y (0,1, j 0,1)} . Next, the 0th layer long-term prediction analysis unit 50d uses this E and the long-term prediction gain γ (0,1) to calculate a long-term residual code amount estimated value PC (0,1) by the following equation.
PC (0,1) = E (1-γ (0,1) ² ) ... (3)

[When the estimated value of the sum of the absolute values of the long-term prediction error amplitude in the block is the long-term residual code amount estimated value]
When the estimated value of the sum of the amplitudes of the long-term prediction errors in the block is used as the long-term residual code amount estimated value, first, the 0th layer long-term prediction analysis unit 50d first selects the short-term prediction residual signal y (0, 1, j _{0 , 1} ) The sum E in the block of the absolute value | y (0,1, j _0,1 ) | Next, the 0th layer long-term prediction analysis unit 50d uses this E and the long-term prediction gain γ (0,1) to calculate the long-term residual code amount estimated value PC (0,1) by the equation (3). . More preferably, the 0th layer long-term prediction analysis unit 50d calculates the long-term residual code amount estimated value PC (0, 1) using the following equation. By using Expression (4), it is possible to calculate the long-term residual code amount estimated value PC (0, 1) with higher estimation accuracy.
PC (0,1) = E (1-γ (0,1) ² ) ^1/2 … (4)
In addition, a long-term filter unit and a long-term prediction coefficient coding unit are provided in the preceding stage, thereby obtaining the sum of the actual energy of the long-term prediction residual signal and the sum of the absolute values of the amplitudes, and calculating the long-term residual code amount estimation value. It is good.

[Processing from the first to the third layer]
Similarly to the 0th hierarchy, the 1st to 3rd hierarchy 1st hierarchy short-term prediction analysis unit 51a, 1st hierarchy short-term prediction coefficient encoding part 51b, 1st hierarchy short-term prediction filter 51c, 1st hierarchy long-term prediction analysis part 51d A first layer residual code amount estimation unit 51e, a second layer short-term prediction analysis unit 52a, a second layer short-term prediction coefficient encoding unit 52b, a second layer short-term prediction filter 52c, a second layer long-term prediction analysis unit 52d, Second layer residual code amount estimation unit 52e, third layer short-term prediction analysis unit 53a, third layer short-term prediction coefficient encoding unit 53b, third layer short-term prediction filter 53c, third layer long-term prediction analysis unit 53d and Each process with the three-layer residual code amount estimation unit 53e is executed. As a result, the short-term prediction coefficient code, the long-term prediction coefficient, the short-term prediction residual signal, and the long-term residual code amount estimation value for each block in each layer are calculated. Each calculated value is associated with information for identifying the corresponding block, and stored in the block table 57b of the memory 17 (see FIG. 12).

[Processing of division determination unit 54]
Next, the comparison unit 54a of the division determination unit 54 reads the long-term residual code amount estimation value corresponding to each block of each layer from the block table 57b of the memory 17 and estimates the residual code amount corresponding to the block of the upper layer. The magnitude of the value and the total value of the long-term residual code amount estimation values respectively corresponding to a plurality of blocks in the lower hierarchy constituting the block are compared. Next, the selection unit 54b uses the comparison result to determine a combination of blocks constituting the frame, and the sum of the long-term residual code amount estimation values corresponding to the respective blocks within the frame is minimized. select. The processing of the division determination unit 54 so far is performed in the same manner as the division determination unit 14 of the first embodiment, for example. Then, the selection unit 54b reads the short-term prediction coefficient code, the short-term prediction residual signal, and the long-term prediction coefficient corresponding to each selected block from the block table 57b (FIG. 12) of the memory 17, and these and the selected block And selection information indicating the combination of.

[Encoding process]
Next, the long-term prediction coefficient encoding unit 56 quantizes the long-term prediction coefficient (the long-term prediction delay amount and the long-term prediction gain) output from the selection unit 54b, respectively (quantized long-term prediction delay amount). And the quantized long-term prediction gain) and output. Further, the long-term prediction coefficient encoding unit 56 calculates and outputs a long-term prediction coefficient code obtained by encoding the quantized long-term prediction coefficient. Next, the long-term prediction filter 58 uses the quantized long-term prediction coefficient output from the long-term prediction coefficient encoding unit 56, and the short-term prediction residual signal output from the selection unit 54b is selected by the selection unit 54b. A long-term prediction residual signal is obtained by filtering every time and output. The filtering here means a process for obtaining a difference between a long-term prediction model (filter) into which a quantized long-term prediction delay amount and a quantized long-term prediction gain are substituted, and a short-term prediction residual signal. In this embodiment, filtering is performed by a primary filter. Then, the residual encoding unit 15 generates a residual code obtained by encoding the input long-term prediction residual signal for each block selected by the selection unit 54b, and outputs the residual code.
The adaptive block length encoding device 50 outputs the residual code output from the residual encoding unit 15, the long-term prediction coefficient code output from the long-term prediction coefficient encoding unit 56, and the selection unit 54b. A short-term prediction coefficient code and selection information are output as a code string.

Next, a sixth embodiment of the present invention will be described. In the sixth embodiment, the short-term prediction analysis and the long-term prediction analysis are used in combination, but unlike the fifth embodiment, the code amount of the long-term prediction residual signal after the long-term prediction analysis is estimated using the short-term prediction analysis result. In the following, items common to the already described embodiments will be described in a simplified manner.
<Configuration>
FIG. 13 is a block diagram illustrating a configuration of the adaptive block length encoding device 60 according to the sixth embodiment. In addition, the same code | symbol as what was used until now is attached | subjected about the part which is common in the Example already demonstrated in FIG.

  As illustrated in FIG. 13, the adaptive block length encoding device 60 according to the sixth embodiment includes a 0th layer short-term prediction analysis unit 50a and a 0th layer short-term prediction coefficient encoding unit 50b that process a signal of a 0th layer block. , The 0th layer short-term prediction filter 50c and the 0th layer residual code amount estimation unit 60d, the first layer short-term prediction analysis unit 51a that processes the signals of the first layer block, and the first layer short-term prediction coefficient encoding unit 51b , First layer short-term prediction filter 51c and first layer residual code amount estimation unit 61d, second layer short-term prediction analysis unit 52a for processing the signals of the second layer block, second layer short-term prediction coefficient encoding unit 52b , Second layer short-term prediction filter 52c and second layer residual code amount estimation unit 62d, third layer short-term prediction analysis unit 53a that processes the signals of the third layer block, and third layer short-term prediction coefficient encoding unit 53 , Third layer short-term prediction filter 53c and third layer residual code amount estimation unit 63d, division determination unit 64, long-term prediction analysis unit 65, long-term prediction coefficient coding unit 66, long-term prediction filter 68, residual A differential encoding unit 15, a control unit 16, and a memory 17 are included. The division determination unit 64 includes a comparison unit 64a, a selection unit 64b, and a calculation unit 64c.

FIG. 14 is a conceptual diagram illustrating the data configuration of the block table 67 b of the sixth embodiment stored in the memory 17. The block table 67b includes a short-term prediction coefficient code Cp (d, u), a short-term prediction residual signal y (d, u, j _{d, u} ) calculated as described later, and a residual code amount estimated value PC (d , U) is associated with information specifying each block B (d, u).

<Processing>
Hereinafter, an adaptive block length encoding method according to the sixth embodiment will be described.
[Block division processing]
First, the same block division processing as that in the first embodiment is performed, and a division table 17 a is generated and stored in the memory 17.
[Processing of Level 0]
Next, the 0-layer short-term prediction analysis unit 10a refers to, for example, block identification information in the partition table 17a of the memory 17, and samples x (0, 1) of the block B (0, 1) in the 0th layer from the partition table 17a. , J _0,1 ) (j _0,1 = 1, ..., n) is read. Then, the 0-layer short-term prediction analysis unit 10a performs a short-term prediction analysis of the read sample x (0, 1, j ₀ , ₁ ) and performs a short-term prediction coefficient α (0, 1, i) (i = 1,... ., p) is calculated and output.

  Next, the 0-layer short-term prediction coefficient encoding unit 50b quantizes the input short-term prediction coefficient α (0, 1, i) and the quantized short-term prediction coefficient α ^ (0, 1, i), and the quantization The short-term prediction coefficient code Cp (0,1) of the short-term prediction coefficient α ^ (0,1, i) is obtained and output. The short-term prediction coefficient code Cp (0, 1) output here is stored in the block table 67b of the memory 17 in association with the information specifying the corresponding block B (0, 1) (see FIG. 14).

Next, the 0th layer short-term prediction filter 50c uses the input quantized short-term prediction coefficient α ^ (0,1, i) and uses the sample x (0,1, j _{0, 1} ) is filtered, and a short-term prediction residual signal y (0,1, j _0,1 ) is obtained and output. The short-term prediction residual signal y (0,1, j _0,1 ) output in this way is stored in the block table 67b of the memory 17 in association with the information specifying the corresponding block B (0,1). (See FIG. 14).

Next, the 0th layer residual code amount estimation unit 60d uses the input short-term prediction residual signal y (0,1, j _0,1 ) and similarly performs the residual code amount estimated value in the same manner as in the first embodiment. PC (0, 1) is calculated and output. The output residual code amount estimated value PC (0, 1) is stored in the block table 67b of the memory 17 in association with information specifying the corresponding block B (0, 1) (see FIG. 14). ).

[Processing from the first to the third layer]
Similarly to the 0th layer, the 1st to 3rd layer first layer short-term prediction analysis unit 51a, the 1st layer short-term prediction coefficient encoding unit 51b, the 1st layer short-term prediction filter 51c, and the 1st layer residual code amount estimation 61d, second layer short-term prediction analysis unit 52a, second layer short-term prediction coefficient encoding unit 52b, second layer short-term prediction filter 52c, second layer residual code amount estimation unit 62d, third layer short-term prediction analysis Each process of the unit 53a, the third layer short-term prediction coefficient encoding unit 53b, the third layer short-term prediction filter 53c, and the third layer residual code amount estimation unit 63d is executed. As a result, the short-term prediction coefficient code, the short-term prediction residual signal, and the residual code amount estimation value for each block of each layer are calculated. Each calculated value is associated with information for identifying the corresponding block and stored in the block table 67b of the memory 17 (see FIG. 14).

[Processing of Division Determination Unit 64]
Next, the comparison unit 64a of the division determination unit 64 reads the residual code amount estimated value corresponding to each block of each layer from the block table 67b of the memory 17, and the residual code amount estimated value corresponding to the block of the upper layer. And the total value of the residual code amount estimation values respectively corresponding to a plurality of blocks in the lower hierarchy constituting the block are compared. Next, the selection unit 64b uses the comparison result to select a combination of blocks that constitute a frame and that has the smallest sum of residual code amount estimation values corresponding to each block in the frame. To do. The processing of the division determination unit 64 so far is performed in the same manner as the division determination unit 14 of the first embodiment, for example. Then, the selection unit 64b reads the short-term prediction coefficient code and the short-term prediction residual signal corresponding to each selected block from the block table 67b (FIG. 14) of the memory 17, and indicates a combination of these and the selected block. Output selection information.

[Encoding process]
Next, the long-term prediction analysis unit 65 performs long-term prediction analysis of the short-term prediction residual signal output from the selection unit 64b for each block selected by the selection unit 64b, and for each block, the long-term prediction coefficient ( (Long-term prediction delay amount and long-term prediction gain) are calculated and output.
Next, the long-term prediction coefficient coding unit 66 quantizes the long-term prediction coefficients output from the long-term prediction analysis unit 65, respectively, and obtains quantized long-term prediction coefficients (quantized long-term prediction delay amount and quantized long-term prediction gain). Calculate and output. The long-term prediction coefficient encoding unit 66 calculates and outputs a long-term prediction coefficient code obtained by encoding the quantized long-term prediction coefficient. Next, the long-term prediction filter 68 uses the quantized long-term prediction coefficient output from the long-term prediction coefficient encoding unit 66 and the short-term prediction residual signal output from the selection unit 64b is selected by the selection unit 64b. A long-term prediction residual signal is obtained by filtering every time and output. Then, the residual encoding unit 15 generates a residual code obtained by encoding the input long-term prediction residual signal for each block selected by the selection unit 64b, and outputs the residual code.
The adaptive block length encoding device 60 outputs the residual code output from the residual encoding unit 15, the long-term prediction coefficient code output from the long-term prediction coefficient encoding unit 66, and the selection unit 64b. A short-term prediction coefficient code and selection information are output as a code string.

<Features of Example 6>
In the sixth embodiment, compared to the fifth embodiment, the number of operations performed by the long-term prediction analysis unit can be reduced by (the number of hierarchies−1) hierarchies. Therefore, the calculation amount can be further reduced as compared with the sixth embodiment. Further, in this modification, a prediction residual code amount estimation value (such as the sum of the short-term prediction residual signal energy or the absolute value of the amplitude in the block) is obtained from the short-term prediction residual signal, and this is calculated as the long-term prediction residual signal. Is used to estimate the amount of code. However, since the effect of reducing the code amount by long-term prediction is almost the same in each channel, the selection by the selection unit 64b is unlikely to be erroneous.

[Other common matters]
Further, in each of the above-described embodiments, the total residual code amount estimated value corresponding to a plurality of blocks in the lower layer and the residual code corresponding to the block in the upper layer composed of the plurality of blocks in the lower layer When the difference from the quantity estimation value is less than or less than a predetermined threshold value, the residual code is calculated for all blocks belonging to the frame, and the residual code corresponding to the upper layer block is calculated for the frame. The code amount is compared with the total value of the code amount of the residual code corresponding to each of a plurality of blocks in the lower layer constituting the block, and is a combination of the blocks constituting the frame. A configuration may be used in which a code that minimizes the sum of the code amounts of the corresponding residual codes in the frame is selected.

  Specifically, for example, the adaptive block length encoding device 70 of FIG. 15 can be exemplified. In this case, under the control of the control unit 16, the code string is usually generated by the same process as in the first embodiment. However, the difference between the total residual code amount estimated value corresponding to the plurality of blocks in the lower layer and the residual code amount estimated value corresponding to the block in the upper layer composed of the plurality of blocks in the lower layer is When the comparison unit 14a (also used as the “determination unit”) determines that the value is equal to or less than a predetermined threshold, the 0th layer residual encoding unit 70e, the first layer residual encoding unit 71e, and the second layer residual The difference encoding unit 72e and the third layer residual encoding unit 73e calculate a residual code for all blocks belonging to the frame thus determined. Then, the comparison unit 14a, for the frame, adds up the code amount of the residual code corresponding to the upper layer block and the code amount of the residual code respectively corresponding to the plurality of blocks of the lower layer constituting the block. And the selection unit 14b selects a combination of blocks constituting the frame, and the sum of the code amounts of the residual codes corresponding to the respective blocks is minimized. .

  Further, the difference between the total residual code amount estimated value corresponding to the plurality of blocks in the lower layer and the residual code amount estimated value corresponding to the block in the upper layer composed of the plurality of blocks in the lower layer is When the threshold code is equal to or less than a predetermined threshold, the residual code is calculated only for the blocks belonging to the determined section, and the section in which the residual code is generated corresponds to the section. The total code amount of the residual code of each layer is compared, and the combination of blocks constituting the section, the sum of the code amount of the residual code corresponding to each block within the section It may be a configuration that selects the one that minimizes.

  Further, the difference between the total residual code amount estimated value corresponding to the plurality of blocks in the lower layer and the residual code amount estimated value corresponding to the block in the upper layer composed of the plurality of blocks in the lower layer is When the threshold value is less than or less than a predetermined threshold, the residual code amount estimation unit calculates a high-order residual code amount estimation value for each block thus determined, and the comparison unit The configuration may be such that the comparison is performed using the next residual code amount estimation value.

For example, when this configuration is realized in the adaptive block length encoding apparatus 10 of FIG. 1, the 0-layer prediction analysis unit 10a, the 1-layer prediction analysis unit 11a, the 2-layer prediction analysis unit 12a, and the 3-layer prediction analysis unit 13a perform prediction analysis. The prediction order used in the above can be changed into two stages of a low-order prediction order p ₁ and a high-order prediction order p ₂ . Then, under the control of the control unit 16, typically performs a process of Example 1 using prediction order p ₁ low order. However, the difference between the total residual code amount estimated value corresponding to the plurality of blocks in the lower layer and the residual code amount estimated value corresponding to the block in the upper layer composed of the plurality of blocks in the lower layer is , when a predetermined a threshold or less, or less than comparison unit 14a (also serves as the "determination unit") is determined, the control unit 16, or again the process of example 1 using higher order prediction order p _2, It performs the process of example 1 using the prediction order p ₂ higher by several frames from the next frame. In addition, the prediction order can be changed in three or more stages, and the same processing may be repeated while updating the prediction order until the difference between the residual code amount estimation values is greater than or equal to a predetermined threshold.

The present invention is not limited to the above-described embodiments. For example, in each embodiment, the absolute value of the amplitude of the prediction residual signal or the value obtained by summing the energy of the prediction residual signal in the block is used as the residual code amount estimation value. However, a value that is monotonically increasing with respect to the absolute value of the amplitude of the prediction residual signal may be used as the residual code amount estimation value. Further, the configurations of the embodiments may be appropriately combined and executed under the control of the control unit 16. Needless to say, other modifications are possible without departing from the spirit of the present invention.
Further, when the above-described configuration is realized by a computer, processing contents of functions that each device should have are described by a program. The processing functions are realized on the computer by executing the program on the computer.

  The program describing the processing contents can be recorded on a computer-readable recording medium. The computer-readable recording medium may be any medium such as a magnetic recording device, an optical disk, a magneto-optical recording medium, or a semiconductor memory. Specifically, for example, the magnetic recording device may be a hard disk device or a flexible Discs, magnetic tapes, etc. as optical disks, DVD (Digital Versatile Disc), DVD-RAM (Random Access Memory), CD-ROM (Compact Disc Read Only Memory), CD-R (Recordable) / RW (ReWritable), etc. As the magneto-optical recording medium, MO (Magneto-Optical disc) or the like can be used, and as the semiconductor memory, EEP-ROM (Electronically Erasable and Programmable-Read Only Memory) or the like can be used.

  The program is distributed by selling, transferring, or lending a portable recording medium such as a DVD or CD-ROM in which the program is recorded. Furthermore, the program may be distributed by storing the program in a storage device of the server computer and transferring the program from the server computer to another computer via a network.

  A computer that executes such a program first stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. When executing the process, the computer reads a program stored in its own recording medium and executes a process according to the read program. As another execution form of the program, the computer may directly read the program from a portable recording medium and execute processing according to the program, and the program is transferred from the server computer to the computer. Each time, the processing according to the received program may be executed sequentially. Also, the program is not transferred from the server computer to the computer, and the above-described processing is executed by a so-called ASP (Application Service Provider) type service that realizes the processing function only by the execution instruction and result acquisition. It is good. Note that the program in this embodiment includes information that is used for processing by an electronic computer and that conforms to the program (data that is not a direct command to the computer but has a property that defines the processing of the computer).

  In each embodiment, the present apparatus is configured by executing a predetermined program on a computer. However, at least a part of these processing contents may be realized by hardware.

  As an industrial application field of the present invention, for example, compression encoding of an acoustic signal can be exemplified.

FIG. 1 is a block diagram illustrating a configuration of an adaptive block length encoding apparatus according to the first embodiment. FIG. 2 is a conceptual diagram illustrating the data configuration of the division table. FIG. 3 is a conceptual diagram illustrating the data configuration of the block table according to the first embodiment. FIG. 4 is a flowchart for explaining an example of processing of the division determination unit. FIG. 5 is a block diagram illustrating a configuration of an adaptive block length encoding apparatus according to the second embodiment. FIG. 6 is a conceptual diagram illustrating the data configuration of the block table according to the second embodiment. FIG. 7 is a block diagram illustrating a configuration of an adaptive block length encoding apparatus according to the third embodiment. FIG. 8 is a conceptual diagram illustrating the data configuration of the block table according to the third embodiment. FIG. 9 is a block diagram illustrating a configuration of an adaptive block length encoding apparatus according to the fourth embodiment. FIG. 10 is a conceptual diagram illustrating the data configuration of a block table according to the fourth embodiment. FIG. 11 is a block diagram illustrating a configuration of an adaptive block length encoding apparatus according to the fifth embodiment. FIG. 12 is a conceptual diagram illustrating the data configuration of a block table according to the fifth embodiment. FIG. 13 is a block diagram illustrating a configuration of an adaptive block length encoding apparatus according to the sixth embodiment. FIG. 14 is a conceptual diagram illustrating the data configuration of a block table according to the sixth embodiment. FIG. 15 is a block diagram showing a modification of the adaptive block length encoding apparatus. FIG. 16 is a block diagram illustrating a configuration of a conventional encoding device. FIG. 17 is a conceptual diagram illustrating the configuration of a frame that is hierarchically divided into blocks. FIGS. 18A to 18C are diagrams showing a modification example when a frame is hierarchically divided.

Explanation of symbols

10-70 Adaptive block length encoding device

Claims (16)

A block dividing unit that generates a plurality of hierarchies composed of one or a plurality of blocks from a frame that is a time interval of an input signal;
A residual code amount estimation unit for calculating a residual code amount estimation value corresponding to each block obtained by the block dividing unit;
When a certain section in the frame is composed of different blocks in two or more hierarchies, a comparison unit that outputs a comparison result of the total value of the residual code amount estimation values of the respective hierarchies corresponding to the section When,
A selection unit that uses the comparison result to select a combination of blocks that constitute the frame, and that minimizes the sum of the residual code amount estimation values corresponding to the blocks in the frame;
A residual encoding unit that encodes a prediction residual signal obtained by predicting an input signal corresponding to each block selected by the selection unit with a first prediction order;
An adaptive block length encoding device comprising: The adaptive block length encoding device according to claim 1,
The residual code amount estimation unit includes:
Calculating the residual code amount estimation value by prediction of a second prediction order that is lower than the first prediction order;
An adaptive block length encoding device characterized by the above. The adaptive block length encoding apparatus according to claim 1 or 2,
The residual code amount estimate is
The absolute value of the amplitude of the prediction residual signal obtained by predicting the input signal of each block with the first or second prediction order or the energy of the prediction residual signal is totaled in each block. Is the value
An adaptive block length encoding device characterized by the above. The adaptive block length encoding apparatus according to claim 1 or 2,
The residual code amount estimate is
A value calculated using the input signal of each block and the prediction coefficient of the first or second prediction order,
An adaptive block length encoding device characterized by the above. The adaptive block length encoding device according to claim 1,
The residual code amount estimation unit includes:
Using a short-term prediction residual signal obtained by short-term prediction of the input signal corresponding to each block with the first prediction order, and a long-term prediction gain obtained by long-term prediction of the short-term prediction residual signal Calculating the estimated residual code amount,
An adaptive block length encoding device characterized by the above. An adaptive block length encoding device according to any one of claims 1 to 5,
When a certain section in the frame is composed of different blocks in two or more hierarchies, the difference in the total value of the residual code amount estimation values of the respective hierarchies corresponding to the section is less than or less than a predetermined threshold A determination unit for determining whether or not
When it is determined that the difference between the total values of the residual code amount estimation values of the respective layers corresponding to the section is equal to or less than a predetermined threshold value,
The residual code amount estimation unit calculates a residual code amount estimation value by high-order prediction for the block corresponding to the section, and the comparison unit uses the residual code amount estimation value to perform the comparison. I do,
An adaptive block length encoding device characterized by the above. An adaptive block length encoding device according to any one of claims 1 to 5,
When a certain section in the frame is composed of different blocks in two or more hierarchies, the difference in the total value of the residual code amount estimation values of the respective hierarchies corresponding to the section is less than or less than a predetermined threshold A determination unit for determining whether or not
When it is determined that the difference between the total values of the residual code amount estimation values of the respective layers corresponding to the section is equal to or less than a predetermined threshold value,
For each block corresponding to the section, a section residual signal encoding unit that encodes each prediction residual signal obtained by predicting the corresponding input signal with the first prediction order and obtains a residual code;
A code amount comparing unit that outputs a comparison result of the total value of the code amounts of the residual codes of the respective layers corresponding to the sections for the section in which the residual code is generated by the section residual encoding unit; ,
Further comprising
The selection unit performs the selection using the code amount of the residual code instead of the residual code amount estimation value for the interval in which the residual code is generated by the interval residual signal encoding unit,
The residual encoding unit encodes a prediction residual signal for a block of which the residual code is not obtained by the section residual signal encoding unit among the selected blocks.
An adaptive block length encoding device characterized by the above. A block division process for generating a plurality of hierarchies composed of one or a plurality of blocks from a frame which is a time interval of an input signal;
A residual code amount estimation process for calculating a residual code amount estimate corresponding to each block obtained by the block division process;
A comparison process of outputting a comparison result of the total value of the residual code amount estimation values of each layer corresponding to each section when a certain section in the frame is configured with different blocks in two or more layers When,
A selection process using the comparison result to select a combination of blocks that constitute the frame, and which has a minimum sum of the residual code amount estimation values corresponding to each block in the frame,
A residual encoding process for encoding a prediction residual signal obtained by predicting an input signal corresponding to each block selected in the selection process with a first prediction order;
An adaptive block length encoding method characterized by comprising: The adaptive block length encoding method according to claim 8,
The residual code amount estimation process is as follows.
Calculating the residual code amount estimated value by prediction of a second prediction order that is lower than the first prediction order;
And an adaptive block length encoding method. An adaptive block length encoding method according to claim 8 or 9, comprising:
The residual code amount estimate is
The absolute value of the amplitude of the prediction residual signal obtained by predicting the input signal of each block with the first or second prediction order or the energy of the prediction residual signal is totaled in each block. Is the value
And an adaptive block length encoding method. An adaptive block length encoding method according to claim 8 or 9, comprising:
The residual code amount estimate is
A value calculated using the input signal of each block and the prediction coefficient of the first or second prediction order,
And an adaptive block length encoding method. The adaptive block length encoding method according to claim 8,
The residual code amount estimation process is as follows:
Using a short-term prediction residual signal obtained by short-term prediction of the input signal corresponding to each block with the first prediction order, and a long-term prediction gain obtained by long-term prediction of the short-term prediction residual signal , A process of calculating the residual code amount estimation value,
And an adaptive block length encoding method. An adaptive block length encoding method according to any one of claims 8 to 12,
When a certain section in the frame is composed of different blocks in two or more hierarchies, the difference in the total value of the residual code amount estimation values of the respective hierarchies corresponding to the section is less than or less than a predetermined threshold A determination process for determining whether or not
When it is determined that the difference between the sums of the residual code amount estimation values of the respective layers corresponding to the section is equal to or less than a predetermined threshold, the block corresponding to the section Calculating a residual code amount estimate by prediction, and performing the above comparison process using the residual code amount estimate;
And an adaptive block length encoding method. An adaptive block length encoding method according to any one of claims 8 to 12,
When a certain section in the frame is composed of different blocks in two or more hierarchies, the difference in the total value of the residual code amount estimation values of the respective hierarchies corresponding to the section is less than or less than a predetermined threshold A determination process for determining whether or not
When it is determined that the difference between the sums of the residual code amount estimation values of the respective layers corresponding to the section is equal to or less than a predetermined threshold value, the block corresponding to the section corresponds to the difference. An interval residual signal encoding process for encoding each prediction residual signal obtained by predicting the input signal with the first prediction order and obtaining a residual code;
A code amount comparison process for outputting a comparison result of the total value of the code amounts of the residual codes of the respective layers corresponding to the sections for the section in which the residual code is generated by the section residual encoding process; ,
Further comprising
In the selection process, for a section in which a residual code is generated by the section residual signal encoding process, the selection is performed using the code amount of the residual code instead of the residual code amount estimation value. And
The residual encoding process is a process of encoding a prediction residual signal for a block in which a residual code is not obtained in the interval residual signal encoding process among selected blocks.
And an adaptive block length encoding method.   A program for causing a computer to function as the adaptive block length encoding device according to any one of claims 1 to 7.   A computer-readable recording medium storing the program according to claim 15.

Images