JPS59153326A - Adpcm decoding circuit - Google Patents

Abstract

PURPOSE:To prevent the deterioration in characteristic from being accumulated by outputting a representative residual signal and both high and low limit residual signals to an output of an inverse adaptive quantizing circuit and correcting a representative decoding signal by the values so as to form a nonlinear PCM decoding signal. CONSTITUTION:When an ADPC code nj is inputted to a terminal 8, the inverse adaptive quantizing circuit 91 outputs a value multiplying the representative value in response to the code nj and a threshold value by a present quantizing width DELTAj. This output signal has a value of a representative residual signal ej corresponding to a residual signal ej of the decoder and a value representing both limits of a section of a signal value represented by the representative residual signal ej, and the larger value is taken as THU and the smaller value is taken as THL. Since an adaptive filter 111 and a fixed filter 112 output a forecast value at present time, a representative decoding signal xj is obtained by adding an output forecast value of the filters 111, 112 by adders 101 and 102 respectively to the representative residual signal ej. The deterioration in characteristic is not accumulated by correcting this signal xj so as to form a nonlinear PCM decoding signal.

Description

[Detailed Description of the Invention] The present invention provides an ADPCM decoding circuit that eliminates the A, D P C ~ 1 circuit, especially the metering noise! Regarding.

Regarding ADPCM, see Proceedings of IEEE published in April 1980, page 488.
For details on page 525, and for improved AI) PCM with characteristics that are strong against transmission line bit errors, see page 198.
'Proceedings published by IEEE in May 2016
of ICASSP 82960IA-pages 963 for details. Hereinafter, the above-mentioned second
AI) PCM is explained based on the literature VC.

Figure 1 shows a conventional A, D P CM code and decoding circuit, which includes an input signal terminal 1, a subtracter 2, an adaptive quantizer 3, an inverse adaptive quantizer 4, an adder 5, and an adaptive predictor. An ADPCM code circuit consisting of a circuit 6 and a code output terminal 7, a code input terminal 8, and an inverse adaptive quantization circuit 9.21f+-
3e-410, an ADPCM decoding circuit consisting of an adaptive prediction circuit 11 and an output terminal 12 is shown.

The adaptive quantization circuit 3 is a circuit that obtains an N-bit length signal smaller than M as an output signal when the input signal is expressed as an M-bit length, and judges the input signal using 2N-1 thresholds. The determination result is output as NHISO1-. In other words, if the quantization width at a certain sample time j is Δ4, and the input signal
The quantization width /'XJ+1 in 1) is companded using the following equation according to the input signal level of the adaptive conversion circuit.

△J＋嘗△zu・M(口j)
(2, but here M(nρ is n,
Table 1 shows an example of the multiplier used when encoding an audio signal sampled at 8 kHz into 4 pinos (N=4).

Table 1 Engineering/IZ, 4 If β is set as a positive constant smaller than 1 in equation (2), as long as the adaptive prediction circuit is a time-invariant filter, the operation of △l will have the effect of leaking the past quantization width. It is known that this makes the transmission line more resistant to bit errors.
ons on Communications J, pages 1362-1365. When the inverse adaptive quantization circuits 4 and 9 receive the ON-bit output signal of the adaptive quantization circuit 3 and the transmitted N-bit quantization circuit output signal, the inverse adaptive quantization circuits 4 and 9 generate an M-bit reproduction input signal corresponding to the threshold value. The output is to predict the transmission signal in reverse. This one. This is called the photonic value.

Since the quantization characteristics shown by equations 0) and (3) have a constant width between threshold values, the width between representative values is also constant, and is called a linear sweetening characteristic. In general, the width between threshold values and the width between representative values are not constant, and they usually have widths that depend on the statistical distribution function of the signal to be quantized, which will be described in detail later.

The transfer functions of adaptive prediction circuits 6 and 11 are the same, and if this is P (Z), then P (Z) = '-aJZ-' (4)
-11. Here, (ajli=1...9k) is called the prediction coefficient at time J, and the predictor input signal at time J is expressed as 4. If the output signal of the reverse prediction device is set to 1, then each coefficient is changed so as to minimize the error.

That is, each coefficient is Here, δ and 2 are positive constants smaller than 1.

The conventional AI) PCM encoding circuit and decoding circuit will be described below with reference to FIG. Human signal sample value X at time J
, is input to the ADPCM encoding circuit from terminal 1,
The subtracter 2 calculates the difference between the input signal '4x' and the output signal '4x' of the adaptive prediction circuit 6, and inputs it to the adaptive quantization circuit 3 as an error signal e1.

As described above, the adaptive quantization circuit 3 converts e into the code entry of N pino I, and at the same time it is output from the terminal 7, it is input to the inverse adaptive quantization circuit 4. The inverse adaptive quantization circuit 4 reproduces an M-bit error signal command from 0. The reproduced error signal command and the output of the adaptive prediction circuit 6 are added by an adder 5 to reproduce the sweetened input signal X. After this,
Adaptation amount The prediction widths of the prediction circuit 3 and the inverse adaptive quantization circuit 4 and the coefficients of the adaptive prediction circuit 6 are modified in order to encode the next input signal as described above. As mentioned above, the coefficient water pressure of the adaptive prediction circuit is modified to minimize the power of the error signal, that is, the power, so the eJ signal has a smaller range than the X signal, and is the same. Considering that bits are encoded, the error generated by adaptive arsonization [circuit 3] becomes smaller as the bit size decreases, and more accurate encoding can be achieved.

On the other hand, in the conventional AI) PCM decoding circuit, the received condensation code n is inputted from the terminal 8, and the inverse adaptive condensation circuit 9 generates a reproduction error signal n.

This e and the output of the adaptive prediction circuit 11 are added together by an adder IO, and X is combined and output to the output terminal 12, and is added to the adaptive prediction circuit 11 for predicting the next sample time. Also on the ADPcM decoding circuit side, adaptive quantization circuit n or error signal command I
From J, the predetermined width of the inverse adaptive quantization circuit is varied occasionally according to equation (2) to minimize the difference between The adaptive prediction circuit 11
The coefficient of is changed according to equation (5).

A I) In the P CMM coding circuit decoding circuit, if the internal states of the inverse adaptive quantization circuits 4 and 9 and the adaptive pre-11+1J circuit 6.11 match, then the ADPCM coding circuit/decoding circuit △ △ ~ path ci 'Xi' xjO values match. Therefore, even if the AI) PCM encoding circuit and decoding circuit are installed at a distance, the human input signal X applied to terminal 1 and the output signal X output from terminal 12 will take almost the same value. . By the way, from terminal 7 of the encoding circuit to terminal 8 of the decoding circuit
Until then, the transmission is lδ, but there is a possibility that a Pino 1 error may occur in the transmission path due to thermal noise or the like. In this case A,
The DPCM decoding circuit often falls into an unstable state and cannot recover. This can be explained as follows.

When the transfer function D (Z) from the output stage of the inverse adaptive quantization circuit 9 to the output terminal 12 of the ADPCM decoding circuit is determined as the transfer function of the adaptive prediction circuit 11 using equation (4), it becomes as follows. As mentioned above, aJ is a value calculated from Kaori, and when a transmission path bit error occurs, the correction value of the prediction coefficient of the prediction circuit adapted to the ADPC'M decoding circuit is ADPCM
The adaptive prediction coefficient of the code circuit is a different value from the prediction coefficient of the first circuit. Equation (6) has poles determined by the prediction coefficients, and as a result of the above transmission line bit error, the position of the pole may go outside the unit circle on the Z plane on the ADPCM decoding circuit side. . In such a situation, the ADPCM decoding circuit goes into an oscillating state and cannot return to normal operation at °C.

(Refer to the second document) In the second document, in order to eliminate this unstable state, equation (6
) is expanded as follows to realize an AI) PCM encoding circuit and decoding circuit having a transfer function excluding poles that move adaptively.

If a value suitable for uniform properties is selected, the above-mentioned truncation error will be small, and the deterioration of encoding quality will be negligible.

Here, the fixed coefficient (the method for calculating the total is detailed in 498 of the above-mentioned first document) according to the average nature of the voice.

Equation f7) A conventional system A1) PCM code same code circuit decoding circuit based on K is shown in FIG. FIG. 2 shows input terminal 1, subtracters 21, 22, adaptive quantizer 3, inverse adaptive iJ quantizer 4, adder 51, 52, . A I)' CM consisting of an adaptive filter 61, a fixed filter 62, and an output terminal 7
Code circuit, input terminal 8, inverse adaptive quantization circuit 9, adder 101, 102, adaptive filter 111, fixed filter 1
12, A I) PCM decoding 1 consisting of output terminal 12
! Consists of I4 road. Fixed filters 62 and 112 are
Using the fixed prediction coefficient (gold) used in equation (4), we have the following transmission function.

Furthermore, the selective filter 61.111 has the following transmission function.

However, each adaptation coefficient is modified as follows, which is e
, the second thing is to be modified in the direction of minimizing the signal power.
It is stated in the literature.

bj-41,1-δ)b! + Inquiry (10)
+ 1 call lj
Now, when the input signal X is input from the terminal 1, the subtracter 21 calculates the difference between it and the output X of the fixed filter 62 to become y, which is input to the subtracter 22. Subtractor 2
In step 2, the output yj of the adaptive filter is subtracted from y and added to the adaptive quantization circuit 3. The adaptive quantization circuit 3 converts ej into forceps and outputs the code nj from the output terminal 7, which is also applied to the inverse adaptive quantization circuit 4 to obtain the drowned error signal e. ej is input to the adaptive filter 61 and used for the filter calculation at the next sampling time, and the output yj of the adaptive filter 61 is added by the adder 51 to reproduce the signal Xj and perform the filter calculation at the next sampling time. Used for calculations. Therefore, if the output of the fixed filter 62 is suitable for the average behavior of the input signal, the amplitude level of the first error signal y decreases, and the output of the adaptive filter 61 is subtracted from this signal. The error signal e of 2 becomes a signal with a further level. Generally speaking, the adaptive prediction circuit 6 in FIG. On the other hand, the adaptive filter 61 in FIG. 2 predicts the next input signal from the error signal, so it has the second
Although the adaptive filter 61 shown in the figure is lower, the fixed filter 6
2 generates a signal related to the characteristics of the average input signal, the encoder of FIG. 2 as a whole can perform encoding comparable to that of the encoder of FIG. 1.

No. Next, the operation of path A, D, P, C6 in FIG. 2 will be explained.

When the collapsing code is input from the input terminal 8, the inverse adaptive quantization circuit 9 reproduces the quantized error signal e, inputs it to the adaptive filter 11, and uses it for the adaptive filter calculation at the next sample time. , and the adder 1019 is checked. y is the output J of the fixed filter 12
J kama, and the human input signal Xj of the signal generator 6111 which is added and quantized by the adder 102
is regenerated and supplied to the output terminal 12 and fixed filter 12. Adaptive filter 111 and fixed filter 12
The transfer functions Pi (Z) and P2 (Z) are as shown in equations (8) and (9), and the transfer function D (Z) from the output of the inverse adaptive quantization circuit 9 to the output terminal 12 is Therefore, it matches Equation (7), and since there is no adaptively moving pole on the Z plane, stable operation can be expected even if a transmission line bit error occurs.

In addition to the above, the ADPCM encoding/decoding circuit may have an adaptive zero/adaptive pole type prediction filter that uses the fixed filters 62 and 112 shown in Fig. 2 to limit the range of pole starvation, but the same applies. ω? , so the details are omitted.

So far, we have looked at the ADPCM code/Tawara code circuit.
Considering the introduction of this ADPCM@ route into an existing PCM network, a format will arise in which a signal encoded with PCM is encoded with ADPCM, and then encoded with PCM again and transmitted, as shown in FIG. 2433 shows a case where two stages of ADPCM encoding/decoding circuits are connected in cascade. As a result, a situation occurs in which I) CM encoding and ADPCM encoding are performed alternately.

Generally, the calculations in the ADPCM encoding/decoding circuit are equivalent to 14 bits when linearizing 8-bit nonlinear PCM, so a 14-bit or more winning code is used because it is necessary to perform encoding on the same level as PCM. is running. For this reason,
ADP
Even if a cascade connection 1e such as a CM code/shun code circuit is made, there is no deterioration due to the connection itself. Therefore, the first ADPCM code/(g
Even if the paired circuits and the following ADPCM code/decoding circuits are connected in cascade, the internal state remains the same! -1 is each ADPCM gradient sign/
Even if the decoding circuit visualizes it in the same way and pastes it in cascade over many F, the result is 0.0 for one stage. ') A D P CM R”l W
; The deterioration remains at I.

However, due to the above-mentioned Hayashi, ADPCM encoding/decoding times W
? r and it is connected by a nonlinear 8 hino) pcMi1 circuit. For this reason, i+
+li; When connected, the number of usable bits decreases, and the connection itself deteriorates due to the fact that the weighting of each bit of the number of n'capacities used is non-linear. This deterioration due to the connection itself is caused even if the initial ADPCM code/decoder circuit and the subsequent ADPCM code/postcode/backward circuit state match at a certain point, the input PCM
The selected AI) PCM codes differ due to the difference in the amount of degradation. Selected ADPCMi publication)'4,
The fact that the number of cases shown in Table 1 given by the adaptive quantization equation (2) is different, and the adaptive predetermined value of equation (5) and equation (g) 1ii
Rather than having different coefficients, the internal states match and become more consistent. For this reason, the deterioration when cascading connections is
In addition to the deterioration caused by the PCM connection, the deterioration caused by the ADPCM encoding/decoding circuit is accumulated by the number of cascade-connected stages, resulting in very large deterioration.

Regarding the mechanism in which the coincidence of the internal states described above collapses,
I EEE 1 as long as the relationship between the threshold value and the representative value of the pre-prediction circuit used in the ADPCM encoding/decoding circuit has the linear quantization characteristics shown by equations (1) and (3).
Proceedings of 197, published in 1979
9ISCAS”, pages 969 to 970, and also uses the property of linear quantization that if the internal states of the box match, the threshold interval and the representative value interval match. A method for maintaining state consistency (see Table 2 of the same document) is also described.

However, conventional internal state maintenance techniques do not allow ADPCM codes/postcodes/backward circuits with nonlinear quantization characteristics, which is commonly practiced to improve the power of quantization n. This nonlinear quantization characteristic is to examine the systematic distribution of the signal input to the quantization circuit and determine the threshold and representative value suitable for this distribution. For example, when the distribution function is Gaussian distribution, If the number of quantization code bits is 4, it cannot be determined as shown in Table 2, according to IRE T published in May 1960.
RansactionsonIrlformatio
n Theory”, pages 7 to 12.

As is clear from Table 2, the inter-threshold interval and the representative value interval do not have a constant width compared to the linear quantization characteristics given by equations (1) and (3). For this reason, the conventional method of maintaining coincidence of internal states using the fact that the threshold interval and the representative value interval are constant cannot be applied, and the ADPCM encoding/decoding circuit having such a quantization circuit can be used as a non-L1! form PC
Cascading through M encoding resulted in accumulation of characteristic deterioration.

The object of the present invention is to
An object of the present invention is to provide an ADPCM decoding circuit in which characteristic deterioration does not accumulate even if the encoding/decoding circuit is subjected to nonlinear PCM encoding.

Another object of the present invention is to provide a method that does not depend on the structure of an adaptive prediction circuit of an ADPCM code decoding circuit or the quantization characteristics of an adaptive quantization circuit (without accumulating characteristic deterioration during cascade connection). .

The ADPCM decoding circuit according to the present invention adaptively decodes the residual signal, which is the difference between the linearized input nonlinear encoded PCM signal and the adaptively predicted prediction signal, at each sample time.
In an ADPCM decoding circuit that receives a coded sequence output from a pre-predicting ADPCM encoder and outputs a nonlinear PCM decoded signal, the encoder A representative residual signal, a low limit residual signal, and a limit residual signal are generated in response to the residual signal at the side, and the quantum signal at the next sample time is generated by the cover prediction code signal. an inverse adaptive quantization circuit that determines the quantization characteristics;
In the representative residual signal from the AfJ inverse adaptive quantization circuit,
Adaptive prediction (to be described later) (additional means for adding code g and generating a representative decoded signal, and non-linear encoding PC for the representative decoded signal)
A nonlinear PCM conversion circuit that performs MK conversion, a linear PCM conversion circuit that linearizes the output of the nonlinear PCM conversion circuit, and an adaptive prediction signal to be described later is subtracted from the output signal of the linear PCM conversion circuit to obtain a representative decoded residual signal. and a subtraction means for generating the representative decoded residual signal, the low limit residual signal and the high limit residual (Ft), the output signal of the non-winning PCM converter is output as is or by +1 or -1. means for adding the signal to a valve assembly type PCM signal No. y;
The representative decoding algorithm or the representative decoding signal and the representative residual signal are input, and an adaptive prediction signal at the current time is generated, and a prediction characteristic at the next sample time is determined. The adaptive inverse t quantization circuit outputs not only the representative residual signal but also the high and low extreme residual signals, and these ifirs are used to generate nonlinearity of the representative signal signal. This is an ADPCM decoding circuit whose general feature is to modify a PCM code to generate a nonlinear PCM code signal.

The present invention will be described in detail below with reference to the drawings. Fourth
The figure shows an embodiment of the ADPCMM path shown in FIG. 2 of the present invention, including an input terminal 8, a dequantization circuit 91,
01 to 104, reduction model] 05 to 106, consisting of an adaptive filter 11, a fixed filter 12, a narrow-nonlinear PCM conversion circuit 120, a nonlinear-linear PCM conversion circuit 121, a comparison circuit 123, a selection circuit 124, and an output terminal 126. Adaptive filter 1], [^1 constant filter 12
, '21 [1 ironware 1 old, 102 is ADPCM in Fig. 2
This is the same as the decoding circuit. Also, linear-nonlinear type 1 P
CM conversion circuit 120, nonlinear-linear PCM conversion circuit 12
1 details in September 1970 Be1l System La
"BSTJ" 1 published by boratories
555311 (explained on page -1588 can be used.
When input, the representative value corresponding to n shown in Table 2,
-Tsukuda and n+1 thresholds are each multiplied by the quantization width Δ, and in this way, the representative value takes a form representative of the interval indicated by both thresholds. If Y( is 8, then n
+] A sufficiently large number (for example, 1oooo) as the threshold value
Establish and use it virtually.

Now, if an ADPCM encoder is input to the terminal 8, the inverse adaptive quantizer path 91 outputs an ADPCM code n.

4. Correspondingly, a value obtained by multiplying the representative value and threshold value shown in Table 2 by the current one prediction width Δ is output. This output signal is a representative residual signal command corresponding to the residual signal ej on the encoder side.
This representative decoding signal command is representative of J.
The values indicate the two limits of the interval of signal values in j, and the larger one is T)I.
U, the smaller one is Tt4L. Adaptive filter 1
1 and fixed filter] 2, the prediction at the current time is 111Φ
・Let the sum be P. ) is being output, so by adding the predicted output values of the adaptive filter 11 and the fixed filter 12 to the representative residual value e, by adding the predicted output values of the adaptive filter 11 and the fixed filter 12, respectively, to the representative residual value e, the representative decoded signal is obtained. Get the intersection. Therefore, the following formula holds.

仝 -all +l), (12
)'' CoJ Here again, the representative decoded signal X is the interval (THL+Pj, T
HU+Pjl) is a representative value.

6th decoding (No. 9 is a linear-nonlinear PCM converter 120
(normal 8 bits) is converted to PCM code X. Furthermore, X is again converted to P by the nonlinear-Gentleman PCM converter
CM Golden conversion signal xR: after △ conversion, decrease η: unit 10
5 and 106, by subtracting the predicted output values of the adaptive filter 11 and the fixed filter 12 at the current time, respectively.
The representative decoded residual signal eR6 is converted and input to the comparator 123.

Therefore, the representative decoded residual signal eR is given by the following equation.

R e, = X, -P,
(13) J j J R.

Now, let's consider a time when food is served within the interval (THL, THU). In this situation, the comparator 123 selects the selection circuit 124.
The PCM code X is selected and output. ADPC in the next stage
If the internal state of the M code circuit is the same as the active internal state, then in the next stage ADPCMq circuit, X is used as a linear input, and e1 is within the interval CTHL, THU). The same ADPCM code as the active one is assigned to the signal. Therefore, the internal state of the active and next-stage ADPCM encoding/decoding circuits is 1b.
becomes.

R・ Next, e, or interval (THL, THU) (, facing) TH
In the case of U: Yes. The input of the active ADPCM code circuit (M"r is also the nonlinear PCM fM signal)
The value obtained by PCM quantization of representative b −5 (IW number ζ is
R in J
Since the relationship between JR S, e and X is shown by formula 03), the interval J
At least one PC in (THL+P,,T)(U+Pj)
It should contain the representative value of M. (If the PCM representative value is not in this interval, the ADPCM that generated this interval
No sign should be selected. ) Furthermore, Ward I[i (
The representative value X in THL+P,, THU+Pj) is
Since the CM covertized value is (x, -x,)~2(--THL-P,), and in this situation, the quantization width of PCM is at most 2 degrees from 1 times the quantization width of ADPCM. It is understood that this happens from time to time. In such a case, the nonlinear PC that generated XR:
The difference between the M code X and the input non-podal PCM code of the active ADPCM code circuit is One thing is clear.

In addition, since the nonlinear PCM code has a special polarity amplitude expression, the comparison circuit 123 activates the selection circuit 124 in this situation, and when X is positive (xR is generated and
MS B of nonlinear PCM hood X (qost 51
gn1f 1cant and it) is 1), X
Addition device 103 'C-+1 (minimum stenob size), when negative (corresponding to when the MSB of X is O), add 9 to X
In order to select the output minus 1 in the device 104,
The input PCM signal of the active ADPCM code circuit and the next stage AD
It will be appreciated that the input signals to the PCMi circuits will be completely equal and that the internal states will remain consistent.

R・ Furthermore, e is not in the interval (THL, THU), that is, XR is not in the interval (THL+P,,THU0P,')j J
Consider the case where Je shuttle<THL. In this case as well, at least one PCM representative value is in the interval [:THL+
It is a lotus that is in P,,THU1P,). Also,
The value of PCM i predicting the representative value X within this interval is
J, the 11-child threshold of PCM is (x,
, THU 10'' It will be understood that this occurs when the quantization width of PCM becomes approximately 1 to 2 times the quantization width of ADPCM. In such a case, the difference between the nonlinear square PCM code X that generated X and the input nonlinear PCM code of the active ADPCM code circuit is 1 for X, which is 74%.
It is obvious that it is a small PCM code.

Therefore, in this situation, the comparison circuit 123 selects the selection circuit 124.
When X is positive (corresponding to when XOMSB force is -1), adder 104 adds -1 to X, and when negative (X
B corresponds to 00 o'clock) In order to quickly select the one added by the adder 103 as the output, the input PCM signal of the active ADPCM code circuit and the input PCM signal of the next stage ADPCM code circuit are combined.
The signals become completely equal, and the operation of the internal state filter is as explained with reference to FIG. 2 as iis & iJJ of the conventional ADPCM.

As described above, according to the present invention, even if the ADPCM code/decoding circuit is connected in multiple stages via the PCM code/decoder circuit, if the internal states of the ADPCM code/decoder circuits match, the ADPCM code/decoder An ADPCM decoding circuit having a characteristic that the characteristics deteriorate only by one stage of the Ishi code circuit can be realized.

Further, although FIG. 4 illustrates the invention in h) for the ADPCM circuit of FIG. 2, it can also be easily applied to the ADPCM circuit of FIG. Furthermore, A in Figure 2
The pre-6111 filter 112 in the DPCM circuit was a fixed filter, but it is an adaptive filter and the output of 91 is converted into 1, (THL-e,), (THU-1,)]
The present invention also applies to a method in which the length from the representative value to the extreme value is set to J3, and the length from the representative value to the extreme value is obtained by subtracting e'B: from x to J3.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a conventional ADPCM code/(31I code circuit). FIG. 2 is a block diagram showing another conventional ADPCM code/decoding circuit. Block diagram No. 41-1 shows the ADPCM signal circuit of the present invention.
It is a good 1. In the figure, 91... Inverse adaptive quantization circuit 111... Adaptive filter 112-... (111 constant filters 101-104... Adders 105-106... 120...h9 type-nonlinear PCM, converter 121
...Nonlinear-linear PCM converter 123... Comparison circuit 124... Selection circuit

Claims (1)

[Claims] Receive a code signal output from an ADPCM encoder that adaptively quantizes a residual signal, which is the difference between a signal obtained by linearizing an input non-linearly encoded PCM signal and an adaptively predicted prediction signal at each sample time. Then, in response to the residual signal on the nonlinear PC side, a representative residual signal, a low limit residual signal, and a high limit residual signal are generated, and at the next sampling time by the lid code signal, an inverse adaptive quantization circuit for determining a spectralization characteristic of the inverse adaptive quantization circuit; an addition means for adding an adaptive prediction signal, which will be described later, to the representative residual signal from the inverse adaptive quantization circuit to generate a representative decoded signal; Non-linear encoding P of the representative decoded signal
a nonlinear PCM conversion circuit for converting into CM, and the nonlinear 1
a linear PCM conversion circuit that linearizes the output of the 3CM conversion circuit; and a subtraction unit that subtracts an adaptive prediction signal, which will be described later, from the output signal of the linear PCM conversion circuit to generate a representative decoded residual signal.
By comparing the representative decoded residual signal, the low limit residual signal, and the high limit residual signal, the nonlinear PCM converter output signal is converted into a nonlinear signal as it is or by adding +1 or -1. means for converting into a PCM decoded signal; 6
11. An adapter that inputs the representative decoded signal or the representative decoded signal and the representative residual signal, generates an adaptive prediction side signal at the current time, and determines the prediction characteristics at the next sample time. The adaptive inverse quantization circuit outputs not only the representative residual signal but also the high and low extreme residual signals, and corrects the nonlinear PCM code of the representative decoded signal according to the value of the output of the adaptive inverse quantization circuit. An ADPCM decoding circuit characterized by a non-linear PCM decoding signal.