JPS5972244A - Adaptive forecasting encoding system - Google Patents

Abstract

PURPOSE:To shorten an encoding and decoding processing time and to improve an S/N ratio by detecting the periodic locality of the residual electric power from a locally decoded residual signal, and adapting the number of quantized bits of the residual signal backward on the basis of the detection result. CONSTITUTION:The difference signal (residual signal) between an input signal X(n) from an input terminal 21 and a forecasted signal X'(n) is generated by a subtracter 22 and quantized by a quantizer 23, whose output is sent as a code sequence I(n) to an output buffer 24. The code sequence I(n) is quantized reversely by a reverse quantizer 25 simultaneously, and an adder 26 adds its output e'(n) and the forecasted signal X'(n) together to input a locally decoded signal X''(n) to a forecasting device 27. The quantization width and number of quantized bits of the quantizer 23 and the coefficient of the forecasting device 27 are adapted on the basis of the decoded residual signal e'(n).

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an adaptive predictive coding system (1) suitable for efficiently encoding audio signals and the like at an information rate of about 16 to 32 kilobits/second, for example.

<Prior art> In conventional adaptive predictive coding, coding efficiency was improved by adapting the predictive coefficients and quantization width of the predictor to the characteristics of the input signal. In order to improve this, a method has been proposed in which the number of quantization bits is allocated unevenly according to the temporal bias (1) of the pseudo residual power.

For example, patent application No. 54-042858, patent application No. 56-177.
564+=The adaptive predictive coding shown uses forward adaptation in which bit allocation processing is performed after detecting the residual power from the input signal.

That is, as shown in FIG. n)
is also supplied to the pseudo residual power calculation unit 14, and the pseudo residual power V+ corresponding to the difference signal (residual signal) between the input signal and the predicted signal in the predictive coding unit 12 is calculated. Then, the residual signal in the predictive encoding unit 12 (2 pairs of quantization widths is adaptively controlled by the quantization width control signal (nl). Also, the pseudo residual power Vj is Bit allocation unit 15ζ - manually generated bit allocation signal b
(n) is calculated, and the coded bits for the residual signal in the predictive encoding unit 12 are adaptively assigned (2). The pseudo residual power Vi is input to the multiplexing unit 13 and the code of the residual signal The transmission line 16 is overlapped with the C output and the transmission line 16.
-・Sent.

Adaptive predictive coding C: using such forward adaptation has the following drawbacks.

(1) Since the pseudo residual power is detected for each fixed frame, a time delay of about 60 milliseconds occurs in the encoding/decoding process.

(2) To transmit power information at the same time as the residual signal (two extra amounts of information are required).

<Summary of the Invention> In order to eliminate the drawbacks of the conventional forward type bit allocation, the present invention detects the periodic locality of the residual power from the locally decoded residual signal, and based on this detects the periodic locality of the residual power. This is adaptive predictive coding characterized by adapting the number of quantization bits of the residual signal to bank words.

<Embodiment> ゛Figure 2 shows an embodiment of the present invention. A difference signal (residual signal) between the input signal x (n,) input from the input terminal 11 and the predicted signal x (nl) is obtained by a subtracter 22, and the residual signal is quantized by a quantizer 23. code sequence I (n)
It is sent to the output buffer 24 as a. At the same time, (two-code sequence processing) the river is inversely quantized by the inverse quantizer 25, and its output e (n,) and the predicted signal x (n) are added together by the adder 26 to produce a locally decoded signal. 0(n) is calculated and input to the predictor 27.The quantization width and number of quantization bits in the quantizer 23, as well as the coefficients of the predictor 27, are calculated using the decoded residual signal ◇(n)(2). It is adapted accordingly.

Next, the adaptation processing of each section will be described. The quantization width adaptation section 28 adapts the quantization width Δ(n) for each sample using the residual signal function (river) according to the following equation.

△(nlz) = △o&'v''(r+) v(n) = βV(n-1) 10(1-β) Q2 (nl
G (I (river) where v(n) is the estimated value of the residual power at the nth sample time, Δ0.β is a constant.G
(I (nl >) is a correction coefficient for the estimated value of residual power during overload, and when the quantization level is the outermost (overload state), G (I (river) = Go ≥ 1, otherwise In this case, G (I (n)) = 1.

△ The number of quantization bits is calculated using the residual signal e (nl) in the buffer section 29 (1), and then using these data in the quantization bit adaptation section 31. The details of the operation will be explained with reference to FIG.

First, Nb residual data e (-N
b), e (-Nb+1),...
- Detect the period of the residual signal using e(-1).

The period T is determined as the delay time τ at which the autocorrelation function of the residual data...τ□, X is maximum. Next (one time axis is divided into subintervals that repeat periodically as shown in Figure 4).The subintervals are equal intervals within one period T (each interval when divided into two,
Assume that these are repeated with a period T. Analysis frame (n = -Nb, -Nb+ 1, ---1)
The position ITd of the subinterval for is determined so that the residual power within the first subinterval (the section to which n, which is T (n) −1 in FIG. 4, belongs) is maximized. At this time, the residual power vi within each partial interval is determined by the following equation.

Next, Nf points in the future (n=o, 1...
. . Nf-1) is found by extrapolating the above-determined subinterval C2. At this time, Nb future residual signals e (0), e (11-・-
- Find the number of quantization bits for e (Nf-1>) using the following formula.

n = 0, 1---=N(.

where R is the average bit rate C bits/sample), C
: is the time length ratio of the subinterval to which Nf future sample points belong. The number of quantization bits is allocated temporally so that the waveform distortion of the decoded signal is minimized.

The predictor 29 is similar to the one used in the adaptive predictive coding method of Japanese Patent Application No. 177,564 (1982) (2. It is composed of a Percoll lattice type filter, and the Percoll coefficients are locally decoded in the predictor adaptation section 32. (Adapted using nl.

As mentioned above, the 1-residual signal has variable bit number b (n3
Therefore, in order to transmit the code sequence at a constant speed, C2 must be transmitted via the output buffer 24. In this case, the total number of bits of the code sequence 1(n) of the residual signal is constant in each frame consisting of N1 samples, so the capacity (length) of the output buffer 24 exceeds the length given by the following equation. Never.

Here, bmax and bmln represent the maximum and minimum values of the number of quantization bits. Further, the upper limit of the delay time at this time is given by (upper limit of buffer length)/(2Rfs) (seconds). Here f5 is the sampling frequency.

On the receiving side, if there is no transmission code error, the same adaptation process as on the transmitting side can be performed using the received residual data. That is, the residual data received in FIG. 36 and is output as a decoded output to the output terminal 37 (2. It is input to the predictor 36 and the prediction adaptation section 38. The prediction adaptation section 38 adds the prediction signal of the predictor 36 The coefficients are adaptively controlled.Inverse quantizer 34
The output of is supplied to the quantization width adaptation section 39 and the buffer section 41, and the inverse quantization width of the inverse quantizer 34 is adaptively controlled by the output of the quantization width adaptation section 39. The quantization bit adaptation unit 42 calculates the number of quantization bits based on the signal from the buffer unit 41 and adaptively controls the number of bits to be inversely quantized by the inverse quantizer 34.The quantization width adaptation unit 39 , the quantization bit adaptation unit 42 and the predictor adaptation unit 38 are the quantization width adaptation unit 28 and the quantization bit number adaptation unit 31 on the transmitting side, respectively.
, and performs the same configuration and operation 1 as the predictor adaptation unit 32.

<Effects> As explained above (2) In this invention, the adaptation processing of the number of quantization bits, quantization width, and prediction coefficients of the predictor is all performed using packed words (2) using the residual data. (
The aforementioned 1977=042858 or 1977-177564
) has the following advantages.

(1) The delay time in the encoding/decoding process is less than 1 millisecond, which can be significantly reduced.

(2) Only the residual signal needs to be transmitted, and the code structure on the transmission path is simplified.

Also, compared to conventional adaptive predictive coding that does not use quantization bit adaptation, the signal-to-quantization noise ratio (S/N ratio) for audio input ≦ 2 is 382 to 4.9 d at 32 kbit/s.
B improve.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an adaptive predictive coding method using conventional quantization bit adaptation; Fig. 2 is a block diagram showing an example of an adaptive predictive coding method based on the present invention (2); FIG. 4 is an explanatory diagram of the partial interval. 21: Input terminal, 22: Subtractor, 23: Quantizer, 25
: Inverse quantizer, 28.39: Quantization width adaptation unit, 29.
41: Buffer, 31.42: Quantization bit adaptation unit,
26: Adder, 27° 36: Predictor, 32, 38: Predictor adaptation unit, 24: Output buffer, 33: Input buffer, 25° 34: Inverse quantizer, 37: Output terminal. Patent applicant Nippon Telegraph and Telephone Corporation

Claims (1)

[Claims] (1) In adaptive predictive coding C2, a means for detecting the periodic uneven distribution of the power of the locally decoded residual signal as a puncture word, and a means for detecting the quantization bit number of the residual signal from these powers over time. An adaptive predictive coding system comprising means for allocating the number of bits non-uniformly and means for adaptively quantizing a residual signal using these quantization bit numbers.