JPS60212044A - Method for coding reflected spectral and coding and decoding circuit - Google Patents

Abstract

PURPOSE:To improve the coding efficiency and to attain simplicity and miniaturization of a circuit by applying independently high efficiency coding to plural resampling signal at the transmission side, transmitting the coded signal with multiplex, receiving a multiplexed signal at the reception side and decoding the signal into plural resampling decoding signal. CONSTITUTION:An input signal before sampling is expressed in Equation I , where F'(omega) is the frequency, F(omega) is a frequency of a signal f'(t) applied to a terminal 1, Tsec is the sampling interval and omega0 is a sampling frequency. The f'(t) consists of an even order sampling signal f'e(t) and an odd number sampling signal f'0(t) by a signal separation switch 41 and each spectrum is expressed in Equations II and III. When the input signal is a voice signal and it is sectioned at several hundred msec, since the f(t) is regarded as a steady-state signal and the f'e(t) and the f0'(t) are handled as steady-state signals. The f'e(t) and f'0(t) outputted from the signal separation switch 41 are inputted respectively to adaptive differential PCM coding circuits 25, 26, they are corrected into coefficients supplying a good forecast value so long as they are steady-state signals, and the f'e(t) and f'0(t) are coded into codes with good coding efficiency.

Description

[Detailed description of the invention] <Industrial application field> The present invention relates to band compression technology for signals such as voice/music.

(Prior art and its problems) As conventional band compression technology, sub-band encoding methods that partially adopt adaptive differential PCM19' encoding method and adaptive differential PC encoding method are well-known, and this article provides an overview of these. As for IE Transactions on Communication'/g7s (I
BFiFi Transactions onComm
710-737.

The adaptive differential PQ coding method and subband coding method will be described below, limited to the necessary range. Quantizer 3,
Inverse quantizer 4, adder 5, predictor 6 10, predictor 11
0 Quantizer 3 is a circuit that obtains an N-bit length output signal smaller than M when the input signal is represented by M bit length as an output signal. The input signal is judged using 2N-1 threshold values, and the judgment result is output in N bits. If the input signal xj at this time is N: the number of allocated quantization bits, the output signal is nj, and at the next sampling time (j
The quantization width Δj+□ at
(nρ J + I J (2, but here M (n J) is a multiplier determined by the coefficient of nj, and the audio signal sampled at 8 kHz is encoded into 4 bits) (N = 4) An example of the multiplier used in this case is shown in Table 1. Hereinafter, mj will be referred to as a quantization step size coefficient.

Table 1 If β is set as a positive constant smaller than l in equation (2), as long as the predictor is a time-invariant filter, the calculation of Δ4 will have the effect of leaking the past quantization width.

It is known that it is resistant to pit errors in the transmission line because of the
Transactions on Communication 7 published by EI)
gns (Transactions on Commun
1 cat tons) Muzzle pages 1362 to 1365
Please take a look at the page. When the inverse quantizers 4 and 9 receive the N-bit output signal of the quantizer 3 and the transmitted N-bit quantizer output signal, the inverse quantizers 4 and 9 calculate M with respect to the threshold value.
It is the one that outputs the bit playback input signal! , -=nj Δj 10.5'J (31, the transmission signal is inversely quantized. The transfer functions of the predictors 6 and 11 are the same, and if this is P(Zl, then )■
=1. ...tk) is called the prediction coefficient at time j, and each coefficient is changed so as to minimize the predictor input at time j. Tsumashi, each coefficient is) + 1 = <-δ>a', 10g-order, A,,
r51 Houou JJ-1 It changes from time to time. Here, δ and g are positive constants smaller than l.

A conventional adaptive differential PCM encoding and decoding method will be described below according to an mx diagram. When the input signal sample value X at time j is input to the adaptive differential P/CM encoder from terminal 1,
Input signal X to subtracter 2.

The difference between J and the output signal XJ of the predictor 6 is calculated and input to the quantizer 3 as an error signal e. As described above, the quantizer 3 converts e into an N-bit code nj, which is output from the terminal 7 and simultaneously input to the inverse quantizer 4. The inverse quantizer 4 generates an error signal e of n, ypM bits.
, play. The reproduced error signal command and the output of the predictor 6 are added by an adder 5 to produce a locally decoded signal X,
Play.

'' Thereafter, the quantization widths of the quantizer 3 and the inverse quantizer 4 and the coefficients of the predictor 6 are modified in order to encode the next input signal as described above. As mentioned above, the coefficients of the predictor are modified to minimize the power of the error signal, i.e., Q-, so that e.

The J signal has a smaller dynamic range than the X signal, and considering that it is encoded using the same bits, the error generated by the quantizer 3 will also be smaller as the J signal is smaller.
On the other hand, in the conventional decoder, the received quantized code nj is inputted from the terminal 8, synthesized by the inverse quantizer 9, and outputted to the output terminal 12. , and is added to the predictor 11 to predict the next sample time.

On the decoder side, the quantization code n, or the error signal
The quantization width of the inverse quantizer is changed from moment to moment, and the coefficients of the predictor 11 are changed so as to minimize the difference between 8 and 7, that is, the number of ej.

In the encoder and decoder, if the internal states of the inverse quantizers 4 and 9 and the predictors 6 and 11 match, the encoder/decoder Q, J, , ? The values of , match. Therefore, even if the encoder and decoder are provided at a distance, the input signal X applied to the terminal l and the output signal xj from the terminal 12 will always take the same value.

This shows a method of dividing into sub-bands, in which a signal input from terminal l is input to a frequency 21 below the sampling frequency and to a low-pass filter 22.

Since the output of the low-pass filter 22 does not contain high-frequency components higher than the sampling frequency, it is possible to reduce the sampling rate to H, and the resampling switch 24 changes the sampling rate to the terminal l. Set the applied signal to H. Similarly, since the output of the high-pass filter 21 is zero in the lower half of the band, the output signal of the high-pass filter 21, which is biased toward the high frequency range, can be resampled to H using the resampling switch 23. , the folded spectrum signal is a folded spectrum signal with a spectrum in which the high-band spectrum is folded back to the low-band side.0Resampling switch The outputs of 23 and 24 are conventional adaptive differential PCs.
The 0 adaptive differential PC input to the adaptive differential PCM encoders 25 and 26 and encoded as described in the M encoding method, respectively.
The codes output from the M encoders 25 and 26 are transmitted from the terminal 7 as one code signal by the multiplexing circuit 27, and become the coding side signal of the sub-band encoder.On the other hand, the sub-band code On the decoding side of the device, the code signal in which the high-band code and the low-band code received through the terminal 8 are multiplexed is sent to the resampling switches 31 and 32, respectively. Separate into area and side codes and apply adaptive differential PC
The signals are input to M decoders 33 and 34, respectively. Adaptive differential PC
The sampled signals decoded by the M decoders 33 and 34 are inserted with a zero value between the sampled signals by the switches 35 and 36, respectively, and the sampling rate is doubled. The details of the properties of the signal with zero inserted between the sampling points are detailed in the first document mentioned above.The frequency spectrum from zero to the sampling frequency is the same as the frequency spectrum before zero is inserted between the sampling points. However, up to the sampling frequency, an inverted version of the spectrum is obtained. Therefore, in the case of the output signal of the switch 35, as described above, the output signal from the resampling switch 23 on the encoder side is Since the adaptive differential PCM decoded signal of the signal obtained by inverting the spectrum of the high-frequency component of the added signal to the low-frequency frequency is inserted between the sample points, the switch 35 The output signal has a spectrum that is approximately the same as the high-frequency component of the signal applied to terminal l up to % of the sampling frequency. Similarly, the output signal of the switch 36 is an adaptation of the sampling rate signal so that the low frequency component of the signal applied to the terminal l is reduced to only the low frequency component by the resampling switch 24 on the encoder side. Since zeros are inserted between sample points for the differential PCM decoding code, the switch 36
The output signal of is from zero to the sampling frequency X at terminal 1.
Therefore, the output of the switch 35 is sent to the high-pass filter 37, and the output of the switch 36 is sent to the low-pass filter 38, respectively. Only the high-frequency and low-frequency components having substantially the same spectrum as the input signal to the input signal 1 are extracted, and the extracted results are added to the adder 39. In this way, a decoded signal of the signal applied to terminal 1 is obtained at terminal 12.

The outline of the conventional adaptive differential PCM method and sub-band coding method has been explained above.The advantage of sub-band coding is that it is efficient when signal components are unevenly distributed in the high or low frequency range; If we set the number of quantization bits of the adaptive differential PCM of the one that is larger than the other one, but keep the overall number of bits the same as when using bidirectional bits, etc.,
Compared to the adaptive differential PCM method alone, local signal components can be encoded better, so better characteristics can be encoded.

By the way, the adaptive differential PCM method was published in 1981 by IEEE.
IEHE Journal of 5solid-8t
It is often realized with a single-chip signal processor as described in ateCircuit magazine, pages 372-376, and in such cases, the larger the bandwidth of the target signal, the fewer the functions that can be realized. 0 An example of realizing adaptive differential PCM with such a single-chip Fugnal processor is the 1982 IE
GLOBECOM 82 CONFERENCE RECORD (GLOBECOMI82CONF dish R band (0-
For signals with 4Kflz (8KHz sampling), adaptive differential PCM using 10 taps of the predictor can be achieved, whereas for signals with twice the bandwidth (0-8 &: 16Kk sampling)
Adaptive difference PC with 2-tap predictor for
However, it cannot be said that sufficient encoding ability can be obtained for wideband signals.

On the other hand, when the sub-band encoder is applied to the 16 Kb sampled signal mentioned above, the adaptive differential PCM encoding and decoding circuit 25, 26.33° 34 in FIG.
Since it only needs to operate with uz sampling, an 8Kh sampling adaptive differential PCM circuit can be used, and in the above-mentioned literature, a 10-tap predictor can be used, which is convenient. A pass-pass filter 22.38 is required. Therefore, these circuits must be added, and the complexity becomes more than twice (approximately three times) that of the encoder.Furthermore, such a filter causes propagation delay,
Applications such as conference calls, where encoded signals transmitted from many speakers are decoded once, the multiple decoded signals are added, and then encoded again (in other words, the encoding and decoding is performed midway). (repeated cases), the propagation delay time accumulates and the delay time becomes large enough to prevent natural conversation.

<Object of the Invention> An object of the present invention is to realize efficient encoding even for wideband signals with simple hardware using a low-band signal encoding circuit. The object of the present invention is to realize a coding circuit that does not cause propagation delay even if a wideband signal coding circuit is configured using a lowband signal coding circuit. , decomposes the sampled and digitized human input signal into a plurality of resampled signal sequences (goods) obtained by resampling each plurality of samples (goods), and decomposes the resampled signal sequences into a plurality of resampled signal sequences (goods) that overlap and do not overlap each other. Each of the sampled signals is independently encoded with high efficiency, and the independently encoded signals are multiplexed and transmitted, and on the receiving side, the independently encoded and multiplexed signals are transmitted. received, decomposes each into independent code signals, performs high-efficiency decoding on each independent code signal to decode into a plurality of resampled decoded signal sequences, and decodes the plurality of resampled decoded signal sequences. Folded spectrum encoding is characterized in that high-efficiency encoding and high-efficiency decoding are performed on decomposed resampled signals from the decoder side to obtain a single decoded input signal by superimposing the decoded signals. method is obtained.

Further, according to the present invention, there is provided a signal separating means for separating a sampled and digitized input signal into a plurality of non-overlapping signals for each sample, and a plurality of signals outputted from the signal separating means (6). ) signals respectively as inputs, and a plurality of high-efficiency encoding circuits that operate on sample times multiplied by multiple exposures, and outputs from the plurality of high-efficiency encoding circuits are predetermined. a transmission-side folded spectrum encoder comprising a code multiplexing circuit that multiplexes and transmits the code in accordance with the order; and a code that separates the code multiplied signal received from the transmission-side folded spectrum encoder into a plurality of code signals. a separation circuit, and a plurality of (
a plurality of M high-efficiency decoding circuits that each receive a plurality of code signals as inputs and operate on sample times multiplied by a plurality of times; and outputs from the plurality of 61 high-efficiency decoding circuits. It consists of at least a reception-side folding decoder consisting of an output multiplexing circuit that multiplexes and outputs according to a predetermined order, and decomposes high-efficiency encoding into a plurality of resampled signals and encodes and decodes them. A folded spectrum encoding/decoding circuit is obtained.

<Example> Similar to the subband encoding method shown in FIG. 2, this embodiment shows an implementation method of the present invention in which a human signal is decomposed into two signals and encoded. In FIG. 3, l is an input terminal, 41 is a signal separation switch, 25 and 26 are conventional adaptive differential PC abbreviation circuits, 42 is a code multiplex selection circuit, 7 is a transmission terminal, 8 is a reception terminal, and 43 is a Code separation switch, 33
and 34 are conventional adaptive differential PCM decoding circuits, 44 is an output multiplex selection circuit, 12 is an output terminal, D, 51 is a transmission side folded spectrum encoder, and 52 is a reception side folded spectrum encoder.

Assuming that the input signal before being sampled is f(t) and its frequency is 9), what signal is applied to terminal l? The frequency component F(ω) of (t) has a sampling interval of T see and a sampling frequency of ω. Then, it becomes. Is this fit) the signal separation switch 41 even sample signal? e(1) and odd sample signal? . (1) Each spectrum is = ``i';, F (a+ -az.) 2n = -x = 2 '7' = 2 Σ (-1)''F (ω once ω.) (812 □22 If the input signal is a speech signal, etc., f (t) can be considered to be a stationary signal if it is divided into several hundred milliseconds for each phoneme, so the equations [7] and (8) Can you do it? e
(g, fo (g can also be treated as a stationary signal. The force output from the signal separation switch 41) and twist are each adjusted by the adaptive difference P.
The 0-adaptive difference PCM encoding circuits 25 and 26 manually input the coefficients of the predictors included in the CM encoding circuits 25 and 26 according to equation (5). The input signals f, (t), fo ('') is modified to a coefficient that gives a good predicted value as long as it is a stationary signal, and the adaptive differential PCM encoding circuits 25 and 26 each use a code with high encoding efficiency.
ett, e 'o('', encode 1.

The outputs of the adaptive differential PCM encoding circuits 25 and 26 are input to the code multiplexing selection circuit 42, and the selection circuit 42 inputs the adaptive differential PCM encoding circuits 25 and 26 into the code multiplexing selection circuit 42. The output code of the differential PCM encoding circuit 25 is outputted to the transmission terminal 7, and the output code of the adaptive differential PCM encoding circuit 26 is outputted to the transmission terminal 7 at odd sampling times.
5, 26, code multiplex selection circuit 42 and transmission terminal 7
The transmission-side folded spectrum encoder 51 consists of the following. The code signal transmitted from the transmission terminal 7 of the transmission-side folded spectrum encoder 51 and received at the terminal 8 is divided into an even code signal and an odd code signal by the code separation switch 43, and each is subjected to adaptive differential PCM decoding. Add to circuits 33 and 34. The adaptive differential PCM decoding circuits 33 and 34 also perform decoding by modifying the coefficients of the internal predictors according to equation (5).

The outputs of the adaptive differential PCM decoding circuits 33 and 34 are the input signals Q of the adaptive differential PCM encoding circuits 25 and 26, respectively.
, (c), signals f8(t) and f(
! t) is decoded and transmitted to the output multiplex selection circuit 44. The output multiplexing selection circuit 44 uses the adaptive difference P at even sample times.
The output signal fJ of the CM decoding circuit 33 and the output signal fJ of the adaptive differential PCM decoding circuit 34 at odd sample times.

'(t) is transmitted to the output terminal 12. Therefore, output terminal 1
2, from equation (7) and equation (81 = (f(tυ (9
) Also, the frequency spectrum is F(ω--ω0)) 9, and a signal almost the same as the signal input to terminal l is obtained. Above terminal 8, code separation switch 4, adaptive difference P
CM decoding circuits 33, 34, output multiplexing selection circuit 44
and the output terminal 12 constitute a reception-side folded spectrum encoder.

The synchronization of the even and odd sample times of the input signal from the terminal l in the transmitter-side folded spectrum encoder 51 and the receiver-side folded spectrum encoder 52 is performed by the code i multiplexing selection circuit 42 that transmits the code signal to the transmission path and the transmission. A predetermined order may be specified between the code separation switches 43 that receive and distribute code signals from the road. For example, if the input signal from terminal l is a 16 I (Hz sampled signal) and the adaptive differential PCM encoding circuits 25 and 26 each output a 4-bit code, then the even sample encoded signal is converted to the previous 4-bit odd sample. If the encoded signal is made up of the last 4 bits, resulting in a total of 8 bits of code word, then the normal P of 64 kbps will be used as the transmission path.
As long as PCM words can be stored together in the CM communication path and the PCM word synchronization is not lost, it is possible to distinguish between even-numbered sample codes and odd-numbered sample codes using a folded spectrum encoder on the receiving side. In the above explanation, only the case where adaptive differential PCM encoding is used has been explained, but it is of course possible to use a stator measurement method instead of using only adaptive differential PCM as a high-efficiency encoding method. .

<Effects of the Invention> As described above, according to the present invention, the input signal sampled at the sampling period T is encoded using a high-efficiency encoding circuit such as adaptive differential PCM that operates at the sampling period 2T. In addition, even when compared with sub-band coding, etc., since high-pass filters and low-pass filters are not required, the overall circuit size can be small, and O that can provide an encoding circuit with no absolute delay due to filters etc.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a conventional adaptive differential PCMf19' encoding method, FIG. 2 is a diagram depicting a conventional sub-band encoding method,
FIG. 3 is a diagram showing an embodiment of the present invention. In FIG. 3, l...input terminal, 41...signal separation switch, 25.26...adaptive differential PCM encoding circuit, 42...code multiplexing selection circuit, 7...transmission terminal% 8...Reception terminal, 43...Code separation switch,
33.34...Adaptive differential PCM decoding circuit, 44...
Output multiplexing selection circuit, 12...output terminal, 51
52 is a transmission side folded spectrum encoding circuit, and 52 is a reception side folded spectrum encoding circuit. 3rd C1 C(2) Figure 2 (b)

Claims (1)

[Claims] 1. On the transmitting side, the sampled and digitized input signal is decomposed into a plurality of M resampled signal sequences that do not overlap with each other and are obtained by resampling each sample. , each of the plurality of resampled signals is independently encoded with high efficiency, the independently encoded signals with high efficiency are multiplexed and transmitted, and on the receiving side, each of the independently encoded signals is encoded with high efficiency. Receive the multiplexed signal, decompose it into independent code signals, perform high-efficiency decoding on each independent code signal, and decode it into a plurality of resampled decoded signal sequences. A folded spectrum encoding method characterized in that four resampled decoded signals are superimposed to obtain one decoded output signal. 2. The sampled and digitized input signal is
signal separation means for separating signals for each of a plurality of non-overlapping quantity samples;
a plurality of high-efficiency encoding means each receiving a plurality of signals as input and operating on sample times multiplied by a plurality of times; and outputs from the plurality of high-efficiency encoding means. A transmitting side folded spectrum encoder comprising a code multiplexing circuit that multiplexes and sends out according to a predetermined order and the transmitting side folded spectrum encoder separate the received multiplexed code signal into a plurality of M code signals. a code separation means; a plurality of high-efficiency decoding means each receiving a plurality of code signals outputted from the code separation means and operating on sample times multiplied by a plurality of M; A folded spectrum comprising at least a receiving side folding decoder comprising an output multiplexing circuit that multiplexes and outputs outputs from the plurality of M high-efficiency decoding means according to a predetermined order. Encoding/decoding circuit.