HK1129169B - Method for reduction of aliasing introduced by spectral envelope adjustment in real-valued filterbanks - Google Patents

Description

Method for reducing aliasing due to spectral envelope adjustment in a virtual filter bank

The patent application of the invention is a divisional application (II) of an invention patent application with the international application number of PCT/EP2003/009458, the international application date of 8/27/2003, the application number of 03822145.4 entering the Chinese national stage and the invention name of a method for reducing aliasing generated by spectral envelope adjustment in a substantial filter section.

Technical Field

The invention relates to a system for adjusting audio signals comprising the use of a spectral envelope of a substantially subband filterbank. Which reduces aliasing generated by spectral envelope adjustments using a substantial subband filterbank. It also allows accurate calculation of the sinusoidal component energies in the substantial subband filterbank.

Background

PCT/SE02/00626 "use of complex modulated filterbanks to reduce aliasing" has shown that complex exponential modulated filterbanks are excellent tools for spectral envelope adjustment of audio. In this process, the spectral envelope of a signal is represented by the energy values corresponding to the channels of a particular filter segment. By estimating the current energy in these channels, the corresponding subband samples are modified to have the expected energy, and thus the spectral envelope is adjusted. If computational complexity constraints prevent the use of complex exponential modulated filterbanks and only cosine modulation (substantial) can be carried out, strong aliasing is obtained when the filterbank is used for spectral envelope adjustment. This is particularly evident for audio signals with strong tonal structures whose aliasing components produce intermodulation with the original spectral components. The present invention provides a solution to this by limiting the increase in value as a function of frequency by a signal dependent method.

Disclosure of Invention

It is an object of the present invention to provide improved techniques for spectral envelope adjustment.

This object is achieved by a signal spectral envelope adjustment device or method according to claim 1 or 19 or according to claim 20.

The present invention relates to the problem of intermodulation due to aliasing in the real filterbanks used for spectral envelope adjustment. The present invention analyzes the input signal and uses the obtained information to limit the envelope adjustment capability of the filterbank by grouping adjacent channel gain values in a sequence determined by the spectral characteristics of the signal at a given time. For a real-valued filter segment, such as a pseudo-mirror symmetric filter where the transition band overlaps only the nearest neighbor, it can show that aliasing is kept below the prototype filter stopband level due to the aliasing cancellation properties. Although this is not an absolute mathematical case, if the prototype filter is designed with sufficient aliasing compression, the filterbank is a perfect reconstruction type from a perceptual point of view. However, if the channel gain of the adjacent channel is changed between analysis and longitudinal effects, the aliasing cancellation properties are violated and aliasing components will be audible in the output signal. By low-order linear prediction of the subband samples of the filterbank channel, the filterbank channel where strong pitch components occur can be estimated by observing the Linear Predictive Coding (LPC) polynomial characteristics. Thus, it is possible to frequency estimate which adjacent channels do not have to have independent gain values to avoid strong aliasing components from being affected by tonal components present in the channels.

The invention comprises the following features:

-subband channel analysis means for evaluating the filterbank channels for the presence of strong tonal components;

-low order linear predictor analysis in the band channel each time;

-gain grouping decisions based on the location of the zeros of the linear predictive coding polynomial;

substantially performed accurate energy calculations.

Drawings

The present invention is now described by way of specific embodiments without limiting the scope or spirit of the invention with reference to the accompanying drawings, in which:

FIG. 1 depicts a frequency analysis of an original signal frequency range containing multiple sinusoidal components covered by channels 15-24 of an M-channel subband filter segment; the frequency resolution of the displayed analysis is deliberately higher than the frequency resolution of the used filterbank to display the sinusoidally occurring filterbank channels;

fig. 2 depicts a gain vector containing the gain values applied to the subband channels 15 to 24 of the original signal.

FIG. 3 depicts the output from the above gain adjustment in an implementation without the spirit of the present invention;

FIG. 4 depicts the output from the above gain adjustment in a complex-valued implementation;

FIG. 5 depicts half per channel where sinusoids occur;

FIG. 6 depicts a preferred channel grouping in accordance with the present invention;

FIG. 7 depicts the output from the gain adjustment described above with the essential implementation of the invention;

FIG. 8 depicts a block diagram of an inventive apparatus;

fig. 9 depicts analysis and synthesis filter segment combinations that may be advantageously used with the present invention.

FIG. 10 depicts a block diagram of the apparatus of FIG. 8 viewed in accordance with the preferred embodiments; and

FIG. 11 is a block diagram of the apparatus of FIG. 8 illustrating gain adjustment in accordance with a preferred embodiment of the present invention.

Detailed Description

The following illustrates the principles of the invention with embodiments that are merely improvements to a substantial filterbank-based spectral envelope adjuster. It is to be understood that other modifications and variations of the devices and details described herein will be apparent to those skilled in the art. It is therefore intended that the present invention be limited only by the claims that will appear and not by the specific details presented herein for the purpose of illustration and explanation of the embodiments.

In the following description, a substantially pseudo-mirror symmetric filter including a substantial analysis and a substantial synthesis is used. However, it should be understood that the aliasing problem proposed by the present invention also occurs in systems with complex analysis and substantial synthesis, and any other cosine modulated filter sections than the pseudo-mirror symmetric filter used in this description. The invention is also applicable to such systems. In a pseudo-mirrored symmetric filter, each channel essentially overlaps only the frequencies of its neighboring neighbors. The frequency response of the channel is shown in the subsequent figures as a dashed line. This is used only to indicate the overlap of the channels and should not be interpreted as the actual channel response given by the prototype filter. In fig. 1, a frequency analysis of the original signal is shown. The figure shows only the frequency range covered by 15 to 25 pi/M of the M-channel filter section. In the following description, the number of channels marked is derived from their low crossover frequency, so that channel 16 covers a frequency range 16 · pi/M to 17 · pi/M outside its neighbor overlap. If there is no modification of the subband samples between analysis and synthesis, the aliasing will be limited by the characteristics of the prototype filter. As shown in fig. 2, if the sub-channel samples of adjacent channels are modified according to a gain vector, the aliasing cancellation characteristics are lost due to the independent gain values of each channel. Thus, as shown in fig. 3, aliasing components will appear in the output signal imaged around the channel intersection region of the filterbank. As shown in FIG. 4, the complex implementation proposed by PCT/SE02/00626, where the output does not interfere with aliasing components, is not true. To avoid severe intermodulation distortion in the output, the present invention teaches that two adjacent channels sharing a sinusoidal component, such as channels 18 and 19 of fig. 1, must be modified identically, i.e., the gain factors applied to the two channels must be the same. This is hereinafter referred to as the coupling factor of these channels. Of course, this means that the frequency resolution of the envelope adjuster is sacrificed to reduce aliasing. Given a sufficient number of channels, however, the loss in frequency resolution is a small penalty in that the severe intermodulation distortion is lost.

To estimate which channels should have a coupling factor, the present invention teaches the use of linear prediction in band. If low-order linear prediction is used, such as second-order linear prediction coding, the frequency analysis tool can resolve a sinusoidal component in each channel. By observing the sign of the first predictor polynomial coefficients, it is easy to determine whether the sinusoidal components are suitable for the upper or lower half of the subband channel frequency range.

Second order linear predictive polynomial

A(z)＝1-α₁z^-1-α₂z^-2 (1)

Is obtained by linear prediction using an autocorrelation method or a covariant method for each channel in a mirror symmetric filter bank that is affected by spectral envelope adjustment. The symbols of the mirror symmetric filterbank channels are defined according to the following:

where k is the channel number and M is the number of channels, and where the frequency reversal of every mirror symmetric filter channel is considered. Thus, each channel may evaluate where strong tonal components are placed, and thus channels sharing strong sinusoidal components may be grouped together. In fig. 5, the symbols of the channels are labeled, so that the sinusoids are placed in the sub-band channel half, where +1 designates the upper half and-1 designates the lower half. The present invention teaches that to avoid aliasing components, the subband channel gain factors should be grouped for channels where channel k has a negative sign and channel k-1 has a positive sign. Thus, the channel symbols depicted in FIG. 5 are given the grouping required according to FIG. 6, where channels 16 and 17 are grouped, 18 and 19 are grouped, 21 and 22 are grouped, and channels 23 and 24 are grouped. This means that the gain values g for the grouped channels k and k-1_k(m) is calculated together, not separately, according to:

where m is at the time point, E_k ^ref(m) is the reference energy, and E_k(m) is the estimated energy. This ensures that the grouped channels obtain phase gain values. The grouping of gain values preserves the aliasing cancellation properties of the filterbank and gives the output according to fig. 7. Here, the aliasing components presented in fig. 3 disappear. If there is no strong sinusoidal component, then the nulls are placed in either half of the z-plane indicated by the channel symbol and the channel will be grouped. This means that no detection based on a determination of whether a strong tonal component is present is required.

In the real-valued filterbank, the energy estimate is not always represented in complex form. If the energy is calculated by summing the squares of the sub-band samples of a single channel, there is a risk of tracking the signal time envelope rather than the actual energy. This is because the sinusoidal components may have any frequency from 0 to the channel width of the filterbank. If the sinusoidal components are present in the filterbank channels, they may have very low relative frequencies, albeit being the high frequency sinusoid in the original signal. If the averaging time is not well chosen for the sinusoidal frequency when in fact the signal energy is practically constant, it is difficult in a real system to estimate the energy of this signal because a vibrato (amplitude variation) may be introduced. However, the present invention teaches that the filterbank channels should be grouped two by two given the location of the sinusoid. This significantly reduces the vibrato problem, as explained below.

In the cosine-modulated filter stage, an analysis filter h_k(n) is a symmetrical low-pass prototype filter p₀The cosine modulated version of (n) is

Where M is the number of channels, k is 0, 1.., M-1, N is the prototype filter order and N is 0, 1.., N. Here, the symmetry of the prototype filter is assumed for N ═ N/2. The following derivation is an example of a half-sample-like symmetry.

Assuming a sinusoidal input signal x (n) Acos (Ω n + θ) with frequency 0 ≦ Ω ≦ π, the subband signal for channel k ≧ 1 may be calculated to be about

Where P (ω) is the transfer prototype filter P₀(N + N/2) substantially discrete time Fourier transform. This approximation is good when P (Ω + π (k +1/2)/M) is small, and this is particularly true if P (ω) is negligible for | ω ≧ π/M, which complies with the discussion assumptions. For spectral envelope adjustment, the average energy within the subband k may be calculated as

Where w (n) is a window of length L. Inserting equation (5) in equation (6) yields:

where ψ (Ω) is the phase term for independent k and W (ω) is the discrete-time Fourier transform of the window. If Ω is close to an integer multiple of π/M, although the input signal is a stable sine, this energy may vary highly. The artificial vibrato type will occur in systems based on the single essential analysis segment band energy estimation.

On the other hand, assuming pi (k-1/2)/M ≦ Ω ≦ pi (k +1/2)/M and P (ω) is negligible for | ω ≦ pi/M, the subband channels k and k-1 have non-zero outputs and these channels will be grouped together as proposed by the present invention. The two-channel based energy estimation is

Wherein

And

for a useful design of the prototype filter, the establishment S (Ω) is approximately fixed in the above frequency range. Furthermore, if window W (n) has a low-pass filter characteristic, | ε (Ω) | is much smaller than | W (0) |, so the energy estimate variance of equation (8) is significantly reduced compared to equation (7).

Fig. 8 depicts an inventive apparatus for signal spectral envelope adjustment. The inventive apparatus comprises means 80 for providing a plurality of subband signals. It should be noted that the subband signal is associated with a channel number k indicating the frequency range covered by the subband signal. The subband signal is generated from a channel filter having a channel number k in an analysis filterbank. The analysis filterbank has a plurality of channel filters, wherein the channel filter having the channel number k has a specific channel response in the overlap range that is overlapped by the channel response of an adjacent channel filter having a low channel number k-1. The overlap occurs in a particular overlap range. As regards the overlap range, reference is made to fig. 1, 3, 4 and 7 for the overlapping impulse responses shown in dashed lines of the adjacent channel filters of the analysis filterbank.

The subband signals output by the means 80 of fig. 8 are input to means 82 for examining the subband signals with respect to aliasing generating signal components. In particular, the means 82 are operative to view the subband signal having an association with channel number k and to view the adjacent subband signal having an association with channel number k-1. This determines whether the subband signal and the adjacent subband signal in the overlap range have an aliasing generating signal component of a sinusoidal component as illustrated in the example of fig. 1. It is to be noted here that the sinusoidal signal components in the subband signal, for example with associated channel number 15, are not placed in the overlap range. The same is true for the sinusoidal signal component in the subband signal having an associated channel number 20. With respect to the other sinusoidal components shown in fig. 1, it is apparent that these are located in the overlap range of the corresponding adjacent subband signals.

The means 82 for viewing is operative to identify two adjacent subband signals having aliasing generating signal components in the overlapping range. Means 82 is coupled to means 84 for calculating gain adjustment values for the adjacent subband signals. In particular, the means 84 operate to calculate a first gain adjustment value for the subband signal on the one hand and a second gain adjustment value for the adjacent subband signal on the other hand. The calculation is performed in response to a positive result of the viewing device. In particular, the computing device is operative to determine that the first gain adjustment value and the second gain adjustment value are not independent of each other but are dependent on each other.

The device 84 outputs a first gain adjustment value and a second gain adjustment value. It should be noted that in the preferred embodiment, the first gain adjustment value and the second gain adjustment value are identical to each other at this point. For example, in the case of modified gain adjustment values already calculated in a spectral band replication encoder, the modified gain adjustment values corresponding to the original Spectral Band Replication (SBR) gain adjustment values are each smaller than the higher value of the original values described later and higher than the lower value of the original values.

The means 84 for calculating the gain adjustment value thus calculates the gain adjustment value for the adjacent subband signal. These gain adjustment values and the subband signals themselves are applied to means 86 for gain adjusting the adjacent subband signals using the calculated gain adjustment values. Preferably, the gain adjustment performed by the means 86 is performed by multiplying the subband samples by a gain adjustment value, so that the gain adjustment value is a gain adjustment factor. That is, a subband signal gain adjustment with several subband samples is performed by multiplying each subband sample from a subband by a gain adjustment factor that has been calculated for the respective subband. Therefore, the exact structure of the subband signals is not touched by the gain adjustment. That is, the relative amplitude values of the subband samples are maintained, while the absolute amplitude values of the subband samples are changed by multiplying the samples by the gain adjustment values associated with the respective subband signals.

At the output of the means 86, a gain adjusted subband signal may be obtained. When these gain adjusted subband signals are input to a synthesis filterbank, which is preferably a real synthesis filterbank, the output of the synthesis filterbank, i.e., the synthesized output signal, does not exhibit significant aliasing components as described above with respect to fig. 7.

It should be noted here that when the gain values of the adjacent subband signals are identical to each other, a complete cancellation of the aliasing components can be obtained. However, when the gain values of adjacent subband signals are calculated to be dependent on each other, at least a reduction of the aliasing components is obtained. This means that an improved aliasing situation has been obtained when the gain adjustment values are not exactly equal to each other but are closer to each other than in the case where no inventive step is taken.

In general, the present invention is used for spectral band replication or High Frequency Reconstruction (HFR) as detailed in WO 98/57436 a 2.

As is known in the art, spectral envelope replication or high frequency reconstruction includes encoder-side specific steps and decoder-side specific steps.

In the encoder, the original signal with full bandwidth is encoded by the source encoder. The source encoder generates an output signal, i.e. an encoded version of the original signal, wherein one or more frequency bands comprised in the original signal are no longer comprised in the encoded version of the original signal. Typically, the encoded version of the original signal contains only the low-band of the original bandwidth. The high band of the original bandwidth of the original signal is not included in the encoded version of the original signal. Furthermore, on the encoder side, there is a spectral envelope analyzer for analyzing the spectral envelope of the original signal in the frequency band that disappears in the encoded version of the original signal. For example, the vanishing band is a high band. The spectral envelope analyzer operates to produce a coarse envelope representation of the frequency bands that are missing in the encoded version of the original signal. This coarse envelope representation may be generated in several ways. A method transmits individual frequency portions of an original signal through an analysis filterbank to obtain individual subband signals corresponding to individual channels in a frequency range, and calculates each subband energy such that these energy values are a coarse spectral envelope representation.

Another possibility is to employ fourier analysis of the vanishing bands and when the speech signal is considered, calculate the energy of the vanishing bands by calculating the mean energy of the spectral coefficients in a group of critical bands grouped as used according to the known Bark scale.

In this example, the coarse spectral envelope representation includes specific reference energy values, where a reference energy value is associated with a specific frequency band. The sbr encoder now multiplexes this coarse spectral envelope representation with an encoded version of the original signal to form an output signal that is transmitted to a receiver or sbr-ready decoder.

A sbr-ready decoder, as known in the art, operates to obtain a decoded version of an original signal by generating a vanishing band using certain or all of the bands obtained by decoding an encoded version of the original signal. Naturally, the decoded version of the original signal also does not contain the vanishing frequency bands. Now, this vanishing band is reconstructed using the bands contained in the original signal by spectral band replication. In particular, one or several frequency bands in the decoded version of the original signal are selected or copied to the frequency band that has to be reconstructed. The exact structure of the copied subband signals or frequency/spectral coefficients is then adjusted using the gain adjustment values calculated using on the one hand the actual energy of the subband signals that have been copied and using the reference energy extracted from the coarse spectral envelope representation that has been transmitted from the encoder to the decoder. Typically, the gain adjustment factor is calculated by determining the share between the reference energy and the actual energy and taking the square root of this value.

This is the case previously described with respect to fig. 2. In particular, fig. 2 shows the gain adjustment values determined, for example, by a gain adjustment block in a high frequency reconstruction or sbr-ready decoder.

The inventive device depicted in fig. 8 can be used to completely replace a typical sbr gain adjusting device or can be used to enhance a prior art gain adjusting device. In a first possibility, gain adjustment values are determined for adjacent subband signals that are dependent on each other in case the adjacent subband signals have aliasing problems. This means that in the overlapping filter response of the filter from which the adjacent subband signals are generated, there are aliasing generating signal components having tonal signal components as discussed in fig. 1. In this case, the gain adjustment value is calculated by means of a reference energy transmitted from a sbr-ready encoder or by means of an estimate of the energy of the replica subband signal, in response to the means for examining the subband signals with respect to aliasing generating signal components.

In other examples where the inventive apparatus is used to enhance the operating capability of an existing sbr-ready decoder, the means for calculating gain adjustment values for adjacent subband signals may be implemented such that it retrieves gain adjustment values for two adjacent subband signals having aliasing problems. Since the aliasing problem is not noticed by typical sbr-ready encoders, the gain adjustment values of the two adjacent subband signals are independent of each other. The inventive means for calculating the gain adjustment value operates on the calculated gain adjustment values of the adjacent subband signals based on the two retrieved "original" gain adjustment values. This can be achieved in several ways. The first method is to make the second gain adjustment value equal to the first gain adjustment value. Another possibility is to make the first gain adjustment value equal to the second gain adjustment value. A third possibility is to calculate the average of the two original gain adjustment values and use this average as the first calculated gain adjustment value and the second calculated gain adjustment value. Another opportunity is to select different or the same first and second calculated gain adjustment values, both lower than the higher original gain adjustment value and both higher than the lower of the two original gain adjustment values. When comparing fig. 2 and fig. 6, the first and second gain adjustment values for two adjacent subbands, which have been calculated to be dependent on each other, are both higher than the original lower value and both lower than the original higher value.

In another embodiment of the present invention in which the provision of subband signal characteristics (block 80 of fig. 8) is performed in accordance with a sbr-ready encoder, the examination of subband signals (block 82 of fig. 8) regarding aliasing product signal components and the calculation of gain adjustment values (block 84) for adjacent subband signals are performed in a sbr-ready encoder that does not require any gain adjustment operations. In this example, the computing means depicted by reference numeral 84 in fig. 8 is connected to means for outputting first and second computed gain adjustment values that are transmitted to the decoder.

In this example, the decoder will receive an indication that the "aliasing-reduced" coarse spectral envelope representation and the aliasing-reduction packets of the preferred adjacent subband signal have been directed. Then, since the gain adjustment values are good, the synthesized signal shows no aliasing distortion, so that no modification is required for a typical sbr decoder.

In the following, a specific implementation of the apparatus 80 for providing subband signals is described. If the invention is implemented in a modern encoder, the means for providing a plurality of subband signals is an analyzer for analyzing the vanishing bands, i.e. the bands not included in the encoded version of the original signal.

If the invention is implemented in a modern decoder, the means for providing a plurality of subband signals is a combination of an analysis filterbank for analyzing a decoded version of the original signal and a spectral band replication means for transposing the low-band subband signals to high-band subband signals. However, provided that the original signal-coded version signal comprises the quantized and potentially entropy-encoded subband signal itself, the providing means does not comprise the analysis filterbank. In this case, the means for providing is operative to extract the entropy encoded and re-quantized subband signals from the transmitted signal input to the decoder. The providing means is further operative to transpose the low-band extracted subband signals to a high-band according to any known transposition criterion as known in the art of spectral band replication or high-frequency reconstruction.

Fig. 9 shows the cooperation of an analysis filterbank (which may be placed in an encoder or decoder) and a synthesis filterbank 90 placed in a sbr decoder. The synthesis filter section 90 disposed in the decoder operates to receive the gain adjusted subband signals to synthesize a high frequency signal, which is then combined with a decoded version of the original signal after synthesis to obtain a full-band decoded signal. Alternatively, the substantial synthesis filterbank may cover the entire original frequency band such that the low-band channel of synthesis filterbank 90 may be provided with a subband signal representing a decoded version of the original signal while the high-band filter channel is provided with the gain adjusted subband signal output by apparatus 84 of fig. 8.

As explained earlier, the inventive calculation of gain adjustment values that are dependent on each other may be a combination of a complex analysis filterbank and a substantial synthesis filterbank, or a combination of a substantial analysis filterbank and a substantial synthesis filterbank, particularly for low cost decoder applications.

Fig. 10 depicts a preferred embodiment of the means 82 for viewing the subband signals. As previously explained with reference to fig. 5, the examining means 82 of fig. 8 comprises means 100 for determining low order predictor polynomial coefficients for the subband signal and the adjacent subband signal to obtain predictor polynomial coefficients. Preferably, the first predictor polynomial coefficients of the second order prediction polynomial defined by equation (1) are calculated as illustrated by equation (1). The device 100 is coupled to the device 102 for determining the sign of the coefficients of the adjacent subband signals. According to the preferred embodiment of the present invention, the determining means 102 operates to calculate equation (2) to obtain the subband signal and the adjacent subband signal. The subband-signal sign obtained by means 102 depends on the predictor polynomial coefficient sign on the one hand and on the channel number or subband number k on the other hand. The means 102 of fig. 10 is coupled to the means 104 for analyzing the symbols to determine a contiguous subband signal having aliasing problem components.

In particular, according to a preferred embodiment of the present invention, the means 104 is operative to determine the subband signals as subband signals having aliasing generating signal components, provided that subband signals having low channel numbers have a positive sign and subband signals having high channel numbers have a negative sign. When considering fig. 5, it is clear that this occurs for subband signals 16 and 17, so subband signals 16 and 17 are determined to be adjacent subband signals having a coupled gain adjustment value. The same applies to subband signals 18 and 19 or subband signals 21 and 22 or subband signals 23 and 24.

Alternatively, it should be noted that another prediction polynomial, i.e. a third, fourth or fifth order prediction polynomial, may be used, and that another polynomial coefficient may be used to determine the sign of the coefficients of the prediction polynomial of e.g. the second, third or fourth order. However, equations 1 and 2 are preferred because they involve low computational expense.

Fig. 11 shows a preferred implementation of the apparatus for calculating gain adjustments for adjacent subband signals according to the preferred embodiment of the present invention. In particular, the apparatus 84 of fig. 8 comprises means 110 for providing an indication of the reference energy of the adjacent sub-bands, means 112 for calculating the estimated energy of the adjacent sub-bands, and means 114 for determining the first and second gain adjustment values. Preferably, the first gain adjustment value g_kAnd a second gain adjustment value g_k-1The same is true. Preferably, the device 114 is operative to perform equation (3) as shown above. It should be noted here that typically the reference energy designation of the contiguous subbands can be obtained from the encoded signal output by a normal sbr encoder. In particular, the reference energy constitutes the coarse spectral envelope information generated by a normal sbr-ready encoder.

The invention also relates to a method of adjusting the spectral envelope of a signal using a filterbank, wherein the filterbank comprises a substantial analysis portion and a substantial synthesis portion, or the filterbank comprises a complex analysis portion and a substantial synthesis portion, wherein if a low channel has a positive sign and a high channel has a negative sign, then the low frequency channel and the adjacent high frequency channel are modified using the same gain value, so that the relationship between subband samples of the low channel and subband samples of the high channel is maintained.

Preferably, in the above method, the gain value is calculated using an average energy of the adjacent channels.

The inventive method of spectral envelope adjustment may be implemented in hardware or software, as appropriate. Such implementation may occur on a digital storage medium, such as a magnetic disk or optical disk with electronically readable control signals, which may be used in conjunction with a programmable computer system to implement the present invention. Generally, the present invention is a computer program product having a program code stored on a machine-readable carrier, which is capable of performing the inventive methods when the computer program product runs on a computer. That is, the present invention is also a computer program having a program code for performing the inventive method, when the computer program product is run on a computer.

Claims (23)

1. An apparatus for calculating gain adjustment values for gain adjusting a plurality of subband signals generated by filtering a signal with an analysis filterbank having subband filters, adjacent subband filters of the filterbank having transition bands that overlap in an overlap range, the apparatus comprising:

means (82) for examining, the means (82) for examining a subband signal originating from a subband filter and an adjacent subband signal originating from an adjacent subband filter to determine whether the subband signal and the adjacent subband signal have aliasing-generating signal components in the overlapping range, thereby obtaining a grouped subband signal in response to the examined positive result;

means (84) for calculating a first gain adjustment value and a second gain adjustment value for the grouped adjacent subband signals, wherein the means for calculating is operative for

A background for estimating energy by determining reference energy with respect to the grouped subband signal and energy in the adjacent subband signal, and

and calculating a gain adjustment value of the grouped subband signal based on the reference energy and the estimated elevation of the energy.

2. The apparatus according to claim 1, characterized in that the apparatus further comprises means (110) for providing a first reference spectral envelope value of the subband signal and a second reference spectral envelope value of the adjacent subband signal;

wherein the means (84) for calculating is operative to determine (112) a first energy measurement and a second energy measurement for the sub-band signal after grouping, the first energy measurement indicating signal energy of the sub-band signal and the second energy measurement indicating signal energy of the adjacent sub-band signal, and

wherein the means (84) for calculating is further operative for calculating (114) an indication of the reference energy from a linear combination of the first and second reference spectral envelope values, and calculating an energy estimate of the energy of the grouped adjacent subband signals from a linear combination of the first and second energy measurements.

3. The apparatus of claim 1 wherein said means (84) for calculating is operative to calculate said first gain adjustment value and said second gain adjustment value such that said first gain adjustment value and said second gain adjustment value differ from each other by less than a predetermined threshold or are equal to each other.

4. The apparatus of claim 3 wherein said predetermined threshold is less than or equal to 6 dB.

5. The apparatus according to claim 1, characterized in that the apparatus further comprises means for providing an unmodified first gain adjustment value for the subband signal and an unmodified second gain adjustment value for the adjacent subband signal, and

wherein the means (84) for calculating is operative to calculate the first gain adjustment value and the second gain adjustment value such that the first gain adjustment value and the second gain adjustment value are both greater than or equal to the lower of the first unmodified gain adjustment value and the second unmodified gain adjustment value and less than or equal to the higher of the first unmodified gain adjustment value and the second unmodified gain adjustment value.

6. The apparatus of claim 5, wherein the unmodified first gain adjustment value and the unmodified second gain adjustment value indicate a spectral envelope of an original signal in a frequency band, wherein the frequency band is reconstructed by spectral band replication.

7. The device according to claim 1, characterized in that said means (84) for calculating is operative for calculating said first gain adjustment value and said second gain adjustment value based on an average energy of said subband signal and said adjacent subband signal.

8. The device of claim 1, wherein the means for calculating is operative to calculate the first gain adjustment value and the second gain adjustment value based on following equations:

wherein g is_k-1(m) is a gain adjustment factor of the first subband signal among the grouped subband signals, g_k(m) is a gain adjustment factor for a second one of the grouped subband signals,is the reference energy of the first subband signal,is the reference energy of the second subband signal, E_k-1(m) is an energy estimate of the first subband signal, E_k(m) is an energy estimate of the second subband signal, and m is a time point.

9. The apparatus of claim 1, wherein the estimate of the energy of the subband signal is calculated by summing the squares of the subband samples of the subband signal.

10. Device according to claim 1, characterized in that the means (82) for examining is operative for calculating signs of subband signals based on prediction polynomial coefficients of the subband signal and the adjacent subband signal (100, 102) and for indicating (104) a positive result when the signs have a predetermined relation to each other.

11. The apparatus of claim 10, wherein the means (82) for reviewing is operative to apply an auto-correlation method or a co-mutation method.

12. The apparatus of claim 10 wherein the prediction polynomial is a low-order polynomial having first-order coefficients, wherein the order of the low-order polynomial is less than 4, and wherein the means (82) for examining is operative to calculate the sign of the subband signal using the first-order coefficients.

13. Apparatus in accordance with claim 10, in which the means (82) for viewing is operative to calculate the sign of the subband signal on the basis of the following equation:

where k is the channel number and α₁Are first order coefficients.

14. An arrangement according to claim 10, characterized in that said predetermined relationship is defined such that said subband signal having associated with said channel number k has a first symbol and said adjacent subband signal having associated with said channel number k-1 has a second symbol opposite to said first symbol.

15. The apparatus of claim 14, wherein the first sign is negative and the second sign is positive.

16. The device of claim 1, characterized in that the means for viewing (82) is operative to perform audio analysis of the subband signal and the neighboring subband signal to determine an audio component of an audio measurement having an audio threshold above.

17. Apparatus in accordance with claim 16, in which the means (82) for viewing is operative to determine whether the audio component is located in an overlap range of channel number k and channel k-1.

18. The apparatus of claim 1, characterized in that the apparatus further comprises a synthesis filter section (90) for filtering the gain adjusted subband signals to obtain a synthesized output signal.

19. The apparatus of claim 18, wherein the analysis filterbank is a real-valued filterbank, and wherein the synthesis filterbank is a real-valued filterbank.

20. The apparatus of claim 18, wherein the analysis filterbank is a complex valued filterbank, and wherein the synthesis filterbank is a real valued filterbank.

21. A method of calculating gain adjustment values for gain adjusting a plurality of subband signals generated by filtering a signal with an analysis filterbank having subband filters, adjacent subband filters of the filterbank having transition bands that overlap in an overlap range, the method comprising:

examining a subband signal originating from a subband filter and an adjacent subband signal originating from an adjacent subband filter to determine whether the subband signal and the adjacent subband signal have aliasing-generating signal compositions in the overlap range, thereby obtaining a grouped subband signal in response to the examined positive result; and

calculating (84) a first gain adjustment value and a second gain adjustment value for the grouped adjacent subband signals, the calculation being performed by,

determining an indication of an energy estimate with respect to a reference energy of the grouped subband signal and an energy in the grouped adjacent subband signal, an

Calculating a gain adjustment value for the grouped subband signals based on the reference energy and the indication of the energy estimate.

22. An apparatus for estimating energy of an signal-based having a secondary channel signal-based that was generated by analysis \28670afilter segment having a subband filter, a neighboring subband filter of the filter segment having overlapping conversion bands in an overlapping range, by an analysis of/28670a, the filter segment having a secondary channel-based signal-based on an entity-pivoting an beam , the apparatus comprising:

means (82) for examining, the means (82) for examining a subband signal originating from a subband filter and an adjacent subband signal originating from an adjacent subband filter to determine whether the subband signal and the adjacent subband signal have aliasing-generating signal components in the overlapping range, thereby obtaining a grouped subband signal in response to the examined positive result; and

means for calculating an energy estimate for the energy of the grouped adjacent subband signals.

23. A method for estimating the energy of a signal having a sub-channel signal generated by filtering the signal with an analysis filterbank having sub-band filters, adjacent sub-band filters of the filterbank having overlapping transition bands in an overlapping range, the method comprising:

examining (82) a subband signal originating from a subband filter and an adjacent subband signal originating from an adjacent subband filter to determine whether the subband signal and the adjacent subband signal have aliasing-generating signal compositions in the overlap range, thereby obtaining a grouped subband signal in response to the examined positive result; and

an energy estimate is calculated for the energy of the grouped adjacent subband signals.