JPH0123971B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to improvements in adaptive quantization and adaptive inverse quantization used for band compression of audio signals and the like. Adaptive quantization is a method that determines the quantization width used for quantization according to the size of the input signal in a timely manner in so-called quantization, in which the sample value of the input signal is expressed using a predetermined number of bits. Input signal sample values can be reproduced with high precision using a small number of bits by adaptive inverse quantization that performs inverse operations. Therefore, if such adaptive quantization and inverse quantization methods are applied to digital transmission of audio signals, the number of bits required per sample can be reduced using quantization and inverse quantization methods that do not have an adaptive function. The amount of information to be transmitted can be reduced, and so-called bandwidth compression can be achieved. Among conventionally known adaptive quantization and adaptive inverse quantization methods, instantaneous adaptive quantization and instantaneous adaptive inverse quantization related to the present invention will be described in detail below. September 1973 for instantaneous adaptive quantization
The principle is as follows, as detailed in pages 1119 to 1144 of BSTJ magazine published by Bell Laboratories. Now, the quantization width at a certain sample time j is Δ _j , and the input signal xj at this time is nj・Δj(nj+1)Δj nj∈{0, ±1, ±2, ..., ±(2 ^m-1 - 1), -2 ^m-1 } m: number of allocated quantization bits (1) If so, the output signal is nj, and the quantization width at the next sampling time is determined as follows. Δ _j+7 = Δj・M(nj) (2) Here, M(nj) is a multiplier uniquely determined by nj, and the 8kHz sampled audio signal is encoded into 4 bits (m=4). Table 1 shows an example of the multiplier used in this case.

[Table] If the quantization width is changed from time to time in this way, if the quantization width is fixed to a constant value, the output signal will always be at a small level of 0, or the output signal will always be overloaded. Even at such a large level,
Not only can the dynamic range of the quantizer be increased by decreasing or increasing the quantization width, but also if the adaptation amount according to equation (2) closely follows the input signal, then (1 ) The quantization determined by the formula can also be performed with high precision. When applying such a quantizer to a transmission device, the quantized and transmitted signal is subjected to the calculation of equation (2) on the receiving side, and the quantization width at each sample time is regenerated to x^ By dequantizing the transmission signal using j=njΔj+1/2Δj (3), it is possible to reproduce a signal x^j having approximately the same magnitude as the input signal xj. The quantization width Δj on the sending and receiving sides is equal, as long as the initial values Δo are equal and the transmission path is guaranteed to have no errors (that is, nj is always transmitted correctly). The multiplier M(nj) by which Δj is multiplied matches and holds true. The instantaneous adaptive quantization method described above is an excellent method for the original purpose of transmitting the input signal with high precision and a small number of bits, but it is important to keep the initial quantization width the same on the transmitting and receiving sides. In addition, from a practical point of view, the quantization width on the transmitting and receiving sides is limited at each sample time, as bit errors occur in normal transmission lines due to line distortion and thermal noise. Situations in which the values are inconsistent frequently occur. How to overcome these difficulties can be found in the 1975 IEEE publication, Transacfions on
As detailed in Communication pages 1362 to 1365, this method replaces equation (2) with the following equation (4). Δ _j+1 = Δj〓・M′(nj) (4) However, β is a number close to 1 but smaller than 1,
M'(nj) is a multiplier uniquely determined by nj, similar to M(nj), and is hereinafter referred to as M(nj) for simplicity.
By modifying as above, equation (4) can be transformed as follows. Δ _j+1 =M(nj)・M(n _j-1 )〓・M(n _j-2 )〓 ²・Δ ₀ 〓 ^j (5
) Further, since β is set to a number smaller than 1, β ^k is k
Asymptotically approaches 0 as becomes larger. From the above, the quantization width on the transmitting side and receiving side is TΔj, respectively.
If we write RΔj, even if the initial values are different (TΔo≠RΔo), the quantization code nj is the same for sending and receiving, so the ratio of quantization widths on the sending side and receiving side at time j is TΔ _j+1 /RΔ _j+1 = (TΔo/RΔo) 〓 ^j (6), which converges to 1 as j increases. In other words, the quantization widths for transmission and reception are equal. Furthermore, when a transmission path bit error occurs, the following can be considered in the same way. If a transmission line bit error occurs at time j', the multiplier M(nj') to be multiplied at the next sample time is different between the transmitting side and the receiving side, resulting in a mismatch in the quantization width. In this case, if we redefine time j' as the initial time and consider that the initial values do not match, it will be understood from the above interpretation that the quantization widths on the transmitting side and the receiving side become equal with time. As described above, the quantization width Δj including the error is raised to the β power with time and approaches 1, so 1.0 is called the quantization reference value. However, conventional algorithms that are resistant to transmission bit errors cannot handle signals with a dynamic range of as much as 60 dB, such as audio signals.
Although this is not a problem when encoding an average signal level, it causes deterioration of quantization accuracy at large or small signal levels. Comparing equations (2) and (3), this is the difference in whether Δj is raised to the β power or not, and can be understood as follows using the graph of Δj versus Δj〓 in FIG. In Figure 1, β=
A graph of 1 means that it is not raised to the β power, and corresponds to the operation corresponding to equation (2). Further, the graph of β=1−α corresponds to the calculation of equation (4). As shown in the figure, both graphs intersect at the quantization reference value of 1.0. Now, consider a case where the signal level is constantly high. In this case, in Fig. 1, Δj often comes close to _ΔL , but in a system that uses β-power, which is resistant to bit errors in the transmission path, _ΔL can be raised to ΔL' by raising to the β- _th power. It can be shrunk. The difference between (Δ _L −Δ _L ′) increases as Δ _L increases, so the next truly necessary quantization width Δ _j+1
When calculating this, it becomes a factor that causes an overload condition. For this reason, encoding accuracy deteriorates for signals with high signal levels. Similarly, when the signal level is constantly low, Δj often comes close to _ΔM in FIG. 1, and in this case, when Δj〓 is performed, it becomes _ΔM ', which becomes a value larger than _ΔM . Therefore, for low-level signals, if the quantization width at the next sampling time is calculated using Δj〓, the adaptability of the quantization width will deteriorate, and the signal sample value will lose digits due to quantization. This is more likely to occur, resulting in deterioration of encoding accuracy. Also, when the value of β approaches 1,
Although deterioration in coding accuracy for high-level or low-level signals can be reduced, the mismatch between the quantization widths on the transmitting side and the receiving side with respect to transmission path bit errors remains unresolved for a long time. The purpose of the present invention is to provide an adaptive quantum system that can eliminate mismatches in quantization widths on the transmitting and receiving sides in a short time in response to bit errors in the transmission path, and maintain high coding accuracy even for high-level and low-level signals. The object of the present invention is to provide a quantization method and an adaptive inverse quantization method. The method of the present invention adaptively determines the quantization width at the next sample time using the quantization reference value, the quantization width and quantization code at the current sample time, and uses this quantization width to In an instantaneous adaptive quantization method that adaptively quantizes and dequantizes a quantized input signal, the quantization reference value is changed from a quantization code at each time at a rate according to a syllable of the input signal. It is characterized by providing means for extracting a quantity and correcting it using the quantity, and adapting the quantization reference value of the conventional instantaneous adaptive quantization method at a speed corresponding to the syllable. The principle of the present invention is to change the graph of β1 so that the intersection point (1.0) between the graph of β=1 and the graph of β1 in FIG. 1 changes according to the syllable level of the input signal. Changing the intersection of these two graphs means changing the quantization reference value, which was conventionally fixed at 1.0. For this reason, it is easier to explain mathematically if Expression (4) is expressed as Expression (7) using the quantization reference value ΔS. Δ _j+1 = (Δj/Δ _S ) 〓 Δ _S M (nj) (7) Obviously, when ΔS = 1.0, it agrees with equation (4), and ΔS≠
When it is 1.0, the intersection of both graphs is moved to ΔS. Furthermore, if the value of _ΔS is fixed, the mismatch between the quantization widths on the transmitting side and the receiving side at the initial time can easily become as shown in Equation (6) even when using Equation (7). It can be derived. Now, set the change range of ΔS as [Δ _S0 , Δ _S1 ],
The rate of change of _ΔS is set to a rate that corresponds to the syllable (a signal whose band is limited to about 100 Hz), and _ΔS is controlled by an amount that depends on the amplitude or energy of the input signal. If _ΔS is changed in this way, the graph corresponding to FIG. 1 will become as shown in FIG. 2. sending side,
Assume that Δ _S is the same on the receiving side, and consider a case where a signal with a constantly large level is being encoded. In this case, the quantization reference value is Δ _S1 ,
The next quantization width will be determined using graph A in FIG. Assuming that the quantization width at time j is ΔL, in the conventional method, graph C in Figure 2
In order to determine the next quantization width using
The deterioration from the case where the graph of β=1 is used is significantly smaller than when the graph is determined using . Similarly, when encoding a low-level signal on a regular basis, graph B in Figure 2 will be used.If we consider the case where the quantization width at time j is _ΔM , then from Figure 2 It will be clear that the quantization width can be determined better than conventional methods. In the following, it will be shown that even such a method has a robust property against transmission path errors. When the quantization reference value at time j is ΔSj, the quantization width at time j+1 is obtained by repeatedly using equation (7) as follows. Δ _j+1 = Δj〓M(nj)ΔSj ^(1- 〓 ⁾ =M(nj)M〓(nj−
1) M〓 ² (nj−2)…M〓 ^j _(o0) ×ΔSj ^(1- 〓 ⁾ ΔSj−1 ^(1- 〓 ⁾ 〓ΔSj−2 ^(1- 〓 ⁾ 〓 ² …ΔS
0 ^(1- 〓 ⁾ 〓 ^j ×Δ0〓 ^j (8) In the case of signals such as voice, there are silent sections, so the quantization reference values of the transmitter and receiver encoders at the beginning of the voiced section are the possible values. It can be expected that the minimum value of ΔS0 is reached. Even if transmission and reception do not have the same quantization reference value, the ratio can be expected to be close to 1. Using equation (8), the quantization width and quantization reference value for the transmitting side and the receiving side are ^T Δj, ^R Δj
If ^T ΔSj and ^R ΔSj are set, the ratio of quantization widths at time j+1 is as follows. However, it is assumed that there is no transmission path error from time 0 to time j. RΔ _j+1 /TΔ _j+1 = (RΔ _Sj /TΔ _Sj ) ^(1- 〓 ⁾ (RΔ _Sj-1 /T
Δ _Sj-1 ) ^(1- 〓 ⁾ 〓(RΔ _Sj-2 ／TΔ _Sj-2 ) ^(1- 〓 ⁾ 〓 ²・(R
Δ _Sj-3 /TΔ _Sj-3 ) ^(1- 〓 ⁾ 〓 ³ ...(RΔ ₀ /TΔ ₀ )〓 ^j (9)
If the value of j is chosen to be around 10, both ^T ΔSj and ^R ΔSj
Since ^T ΔS0 and ^R ΔS0 are considered to be almost equal,
(The rate of change of the quantization reference value is about 100Hz, that is,
When performing 8KHz sampling, one period of 100Hz is
There are 80 samples, and there is no significant change in the 10 sample interval we are currently considering. ) Equation (9) becomes as follows. RΔ _j+1 ／TΔ _j+1 = (RΔ _S0 ／TΔ _S0 ) ^(1- 〓 ⁾⁽¹⁺ 〓 ⁺ … ⁺ 〓 ^j
) = (RΔ _S0 /TΔ _S0 ) ^1- 〓 ^j+1 ≒RΔ _S0 ／TΔ _S0 ≒1(10) Therefore, there is almost no difference in signal level between transmitting and receiving,
This is the difference between the quantization reference values that occur at the beginning of speech. Regarding the change in the quantization reference value when a transmission line bit error occurs, only a slight difference appears as shown below. Assume now that a transmission path bit error occurs at time j, and the quantization codes on the transmitting side and the receiving side are different. Let us consider the influence of these different quantization codes on the quantization reference value, which changes at a speed corresponding to the syllable. The decoded signal at time j becomes discontinuously larger than the input signal by an amount corresponding to the sign of the difference caused by the transmission line bit error.
The energy of the difference between the decoded signal from this discontinuous point and the input signal becomes the difference between the quantization reference values. but,
The discontinuity points generally have many high-frequency components and are impulse-like, and the quantization reference value is the energy of the decoded signal, that is, the output of a low-pass filter having a cutoff characteristic at 100 Hz, as described later. Therefore, 4K of audio etc.
In an 8KHz sampling system that handles Hz band-limited signals, this impulse signal appears on the quantization reference value as an influence of (100Hz/4000Hz) ² = 1/1600 (11). Therefore, the influence on the quantization reference value can be ignored unless it is a long-term error such as a burst error. Furthermore, such errors do not accumulate as long as signals such as audio are handled, and the quantization reference values of both the transmitter and the receiver become the predetermined minimum value ΔS0 again in the silent section of the audio and match. In any case, unless a transmission line bit error causes a negligibly small error in the quantization reference value in the transmission and reception, the quantization width in the transmission and reception is the same as in the conventional quantization algorithm. The error is attenuated moment by moment due to the β power effect of equation (4) or equation (7). Next, the present invention will be explained in detail with reference to the drawings. FIG. 3 shows an embodiment of the present invention, in which 1 is an input terminal for a sampled and digitized signal, 2 is a divider that takes the input signal as the dividend, and 3 is a quantization circuit that quantizes the output of the divider. , 4 is a quantization code output terminal, 5 is a quantization reference generation circuit that outputs a quantization reference value that changes from the quantization code at a rate according to the syllable, and 6 is a quantum that becomes the divisor of the divider 2 at the next sampling time. It consists of a quantization width determining circuit that calculates the value of the quantization step width. Although the details of the quantization circuit 3 will be described later, a plurality of threshold values (for example, 2 ^m -1) are provided for the digitized input signal, and the threshold value is expressed by a smaller number of bits (for example, m) than the number of input signal bits. This converts the signal into an output signal. The details of the quantization reference generation circuit 5 will be described later, but it calculates the input signal energy from the output signal and outputs a quantization reference value according to a change in the input signal energy. The size of the quantization width depends on the quantization code. Furthermore, the quantization code depends on the prediction residual signal. Therefore, when the prediction closely matches the input signal, the quantization width becomes very small. Unfortunately, while the predictions are accurate to some extent, they are never exact. This can be seen, for example, in 215 of ``Digital Processing of Speech Signals,'' published by Prentice Hall, Inc., 1978.
As described on the page, the prediction gain (improvement in S/N by using prediction) is at most 10 dB.
That's about it. Therefore, when the input signal energy is large, the prediction residual signal energy becomes large, and as a result, a large quantization width is likely to occur. The conventional quantization reference value is set so that the quantization reference value is 1 for an average input signal. Furthermore, in the design of the device, it is desired to use the average input signal energy as a reference for the input signal, so it is set to 1. Therefore, the present invention uses the input signal energy as a quantization reference value. The details of the quantization width determination circuit 6 will be described later, but the next sample is determined based on the quantization width that is the output of the quantization width determination circuit at the current time, the output of the quantization reference generation circuit 5, and the quantization code at the current time. This is a circuit that determines the quantization width at time. Now, assume that at a certain time j, xj is applied to the input terminal 1, and the quantization width determining circuit 6 outputs the quantization width Δj at the current time. At this time,
The output of the divider 2 is xj/Δj. Therefore, the quantization circuit 3 outputs from the terminal 4 an m-bit code nj such that njxj/Δj<(nj+1) (11). In this case, equation (11) can be transformed as nj·Δjxj<(nj+1)Δj (12) and thus matches equation (1). Quantization circuit 3
The output nj is applied to the quantization reference generation circuit 5, which updates the calculated signal energy value ΔSj _-1 by a very small amount and changes the quantization reference value ΔSj according to the updated amount of energy, as described later. let The quantization width determination circuit 6 outputs the output nj of the quantization circuit 3, the output ΔSj of the quantization standard generation circuit 5, and the quantization width determination circuit 6.
Using the output Δj, calculate Δ _j+1 = Δj 〓M(nj) ΔSj ^(1- 〓 ⁾ (13) and determine the quantization width Δ _j+1 at the next sampling time. The reason why the adaptive quantization circuit shown in Figure 3 is resistant to transmission line bit errors and can perform highly accurate quantization is as follows.
The output of the quantization circuit 3 can be determined by formula (12), which is equal to formula (1), the quantization width at the next sampling time can be determined by formula (13), which is equal to formula (8), and the above-mentioned This will be clearer from what has been detailed in the principle section of the invention. Next, the configuration of the quantization circuit 3 used in FIG. 3 will be described. In the following, for the sake of simplicity, a case will be described in which an input signal equivalent to 4 bits is quantized into an output signal of 2 bits. The quantization circuit 3 can use a read-only memory (ROM). The 4-bit input signal is
Assuming that the signal is supplied to the lower 4 bits of the address input terminal of the ROM and a 2-bit output signal is output from the lower 2 bits of the ROM output terminal, the contents of the ROM for configuring the quantization circuit 3 are as shown in Table 2. It becomes like this. However, the input signals added to the address are the 2nd and 3rd bits from LBS as signal values.
It is assumed that there is a decimal point between the bits, and the output of the quantization circuit is considered to be rounded down to an integer value.

[Table] FIG. 4 shows the structure of the quantization width determining circuit 6 of FIG. 3. In FIG. 4, the quantization width determining circuit 6 is
From a quantization code input terminal 11, a quantization width reference input terminal 12, a current quantization width input terminal 13, a multiplier generation circuit 14, a divider 15, a β power calculation circuit 16, multipliers 17 and 18, and an output terminal 19. It is configured. Divider 15 has the same configuration as divider 2 in FIG. Furthermore, the multiplier generating circuit 14 can use the ROM used in the quantization circuit 3 shown in FIG. Set to the value of Furthermore, the β power calculation circuit is also
ROM is available. In this case, decimal numbers are also allowed for the input signal, so if the format is such that there is a decimal point on the lth bit from the LSB, the value obtained by multiplying the input signal by 2 ^l and making it an integer is the value of the input signal raised to the β power. Just store it. Now, when the quantization width Δj at time j is input from the terminal 13 and the quantization width reference signal _ΔSj is input from the terminal 12, the divider 15 calculates Δj/ _ΔSj and outputs it. This signal is applied to the β-power calculation circuit 16, and a signal (Δj/ΔSj) is obtained as an output. This signal is multiplied by _ΔSj added from the terminal 12 by the multiplier 17, resulting in (Δj/ _ΔSj )〓· _ΔSj . On the other hand, the quantization code nj input from terminal 11
is input to the multiplier generating circuit 14 and becomes M(nj),
is applied to multiplier 18. The multiplier 18 multiplies this signal by the output signal of the multiplier 17, and as a result, the output terminal 19 receives (Δj/ _ΔSj )〓・_ΔSj・M
(nj) is output. This result agrees with equation (13),
It will be clear that the circuit of FIG. 4 performs the operation of the quantization width determination circuit 6. FIG. 5 is a block diagram of the quantization reference generation circuit 5 used in FIG. In the figure, the quantization standard generation circuit 5 includes a quantization code input terminal 20, a quantization width input terminal 21, an inverse quantization circuit 22, multipliers 23 and 24, a digital low-pass filter 25,
It is composed of an output terminal 26. In FIG. 5, multipliers 23 and 24 are multipliers 1 in FIG.
The same hardware as 7 and 18 is available. Details of the configuration of the digital low-pass filter 25 can be found in the 1975 Prentice-Hall, Inc., Englewood
“Theory and
You can refer to Fig. 5.10 on page 306 of “Application of Digital Signal Processing”. Inverse quantization circuit 2
2 can use the same ROM as the quantization circuit 3 in FIG. 3, but its contents are different. Table 3 shows the ROM contents of the inverse quantization circuit corresponding to Table 2. However, the contents of the ROM have decimal points at the second and third bits from the LSB. As a result, (nj+1/2) is output for input nj, reproducing the value obtained by dividing equation (3) by Δj.

[Table] Now, consider the case where the quantization code nj is input to the input terminal 20 in FIG. The numerical expression precision (number of bits) of the quantization code nj is increased by the inverse quantization circuit 22, and this signal is processed by the multiplier 23 as follows.
It is multiplied by the quantization width Δj applied from the terminal 21 to reproduce the input signal xj. This reproduced signal xj is squared by a multiplier 24 and passed through a low-pass filter 2.
{x ² _i | i=0, which was previously input by 5.
1,...P} is extracted. In this way, the output of passing the square value of the input signal through the low-pass filter is
It represents the short-term energy of the signal. The output signal of this filter 25 is ΔSj. As described above, according to the present invention, it is possible to realize an adaptive quantizer that is resistant to transmission path bit errors and capable of generating quantized codes with high precision. The fixed quantization circuit 3 shown in FIG. 3 used to explain the present invention realizes a linear quantization circuit that satisfies formula (1) according to Table 2, but the fixed quantization circuit 3 shown in FIG. quantization accuracy can be further improved by using a nonlinear quantization circuit. in this case,
Non-linear quantization is possible by changing the number of consecutive pieces of data that have the same content in the ROM content shown in Table 2. Even if such a change is made, as is clear from the principle of the present invention described above, the advantages of the present invention, such as being strong against transmission line bit errors and performing quantization with high accuracy, will not be lost. Therefore, such changes are also part of the invention. The level detector shown in FIG. 5 used in the present invention calculates energy by regenerating the input signal from the quantization code nj using an inverse quantization circuit, but as the signal level increases, the quantization circuit Adaptive quantization circuits tend to be overloaded and also tend to lose digits as the signal level decreases, which also occurs for a short period of time, so the quantization code nj is immediately squared. However, it is also possible to pass the signal through a low-pass filter. In this case, the inverse quantization circuit 22, multiplier 23 and terminal 21 shown in FIG. 5 are omitted. Furthermore, instead of squaring the signal and passing it through a low-pass filter, it is also possible to calculate the energy by passing the absolute value of the signal through a low-pass filter. Even with the above-mentioned various changes, the advantages of the present invention, which are strong against bit errors in the transmission line and highly accurate quantization, remain unchanged. Next, the adaptive inverse quantization circuit of the present invention will be explained with reference to FIG. The adaptive dequantization circuit shown in FIG.
a multiplier 36 whose multiplicand is the output of the quantization code, a quantization reference generation circuit 34 which generates a quantization reference value from the input quantization code, the input quantization code, the quantization reference value described later, and the multiplier of the multiplier 36. (quantization width)
Accordingly, the multiplier (quantization width) of the multiplier 36 at the next sampling time is determined. Quantization width determination circuit 35
and a quantization reference generation circuit 34 that changes the quantization code at a speed corresponding to the syllable. The dequantization circuit 31 is the same as the dequantization circuit 22 shown in FIG. 5, and the circuit shown in FIG. 4 can be used as the quantization width determination circuit 35. Further, the quantization standard generation circuit 34
The circuit shown in the figure can be used. Furthermore, the multiplier 36
is the same as the multiplier 23 in FIG. Now, when the quantization code nj is input to the terminal 30,
Expression accuracy (number of bits) is improved by the inverse quantization circuit 31.
is converted to a higher value and multiplied by the quantization width Δj at this time. From the explanation of FIG. 5, since the output of the quantization circuit 31 was (nj + 1/2), (nj + 1/2) Δj is output to the output terminal 32, and x^j of equation (3) is output. It turns out. On the other hand, in the quantization width determination circuit 35, Δj input from terminal A, nj input from terminal B, and terminal C
The quantization width to be used at the next sampling time is calculated from the input ΔSj by the following calculation: (Δj/ _ΔSj )〓· _ΔSj ·M(nj), and is output from the terminal D. In addition, the quantization standard generation circuit 34 also generates a quantization code.
x^j is reproduced from nj and the quantization width Δj, the energy value of the input signal is updated, and the result is output as the quantization standard ΔSj. The fact that the adaptive inverse quantization circuit having such a configuration is resistant to the influence of bit errors on the transmission line and can perform inverse quantization with high accuracy is detailed in the above description of the principles of the present invention. Note that also in the adaptive inverse quantization circuit,
As described in the configuration of the adaptive quantization circuit, changes such as changing the quantization circuit 31 to a nonlinear inverse quantization circuit and simplifying the calculation of the quantization reference generation circuit 34 can be applied. Further, adaptive quantization and inverse quantization can be used in combination with various other band compression methods, and the use of the principles of the present invention even in the use of such a combination is part of the present invention. An example of such a combination is differential encoding, in which a predictor is used to predict the sample value of the next sample, and the difference between the input signal and the predicted result is quantized and transmitted. It is also part of the invention that uses the principles of the invention. Further, similar effects can be obtained when the adaptive quantization and adaptive inverse quantization circuits of the present invention are implemented as software for a signal processing microprocessor, and are a part of the present invention.

[Brief explanation of drawings]

FIG. 1 is an explanatory diagram of the conventional adaptive quantization and inverse quantization method, FIG. 2 is an explanatory diagram of the adaptive quantization and adaptive inverse quantization method of the present invention, and FIG. 3 is an explanatory diagram of the adaptive quantization and inverse quantization method of the present invention. 4 is a detailed diagram of the quantization width determination circuit 6 of FIG. 3, FIG. 5 is a detailed diagram of the quantization reference generation circuit 5 of FIG. 3, and FIG. FIG. 2 is a block diagram showing a converter. In FIG. 3, 2...divider, 3...quantization circuit, 5...quantization reference generation circuit, and 6...quantization width determination circuit. In FIG. 6, 31...inverse quantization circuit, 34... quantization reference generation circuit, 35... quantization width determination circuit, 36... multiplier.

Claims (1)

[Claims] 1. From the quantization reference value ΔS _j and the multiplier M(n _j ) determined by the quantization width Δ _j and quantization code n _j at the current sampling time (j), the next sampling Time (j+1)
_The quantization _{width Δ j +1 in}
1) In an adaptive quantization and dequantization method that adaptively performs quantization and dequantization using this quantization width, the dequantized signal obtained by dequantizing the quantization code n _j An adaptive quantization and inverse quantization method characterized in that the energy of is determined, and when this energy is large, the above-mentioned ΔS _j is made large, and when this energy is small, the above-mentioned ΔS _j is made small. 2 From the quantization reference value ΔS _j and the multiplier M(n _j ) determined by the quantization width Δj and quantization code n _j at the current sampling time (j), calculate the next sampling time (j+1)
_The quantization _{width Δ j +1 in}
1) An adaptive quantization and inverse quantization device that adaptively performs quantization and inverse quantization using this quantization width determined by The energy of the input signal is determined from the output signals of the divider that divides the signal, the quantization circuit that quantizes and outputs the output signal of the divider, and the output signal of the quantization circuit up to the present, and when this energy becomes large, the ΔS _j From the output of the quantization reference generation circuit, the quantization width and the quantization code at the current time, the quantization reference generation circuit outputs the quantization reference generation circuit that increases ΔS _j and reduces the ΔS j when it is smaller, and the quantization width and quantization code at the current time. An adaptive quantization circuit comprising at least a quantization width determining circuit that determines a quantization width, and an inverse quantization circuit that inversely quantizes an input quantization code, and an input signal of the adaptive quantization circuit from the input quantization code. From the quantization reference generation circuit that calculates the energy and outputs the calculation result as the quantization reference value, the output of the quantization reference generation circuit, the quantization width at the current time, and the input quantization code, the value at the next sampling time is calculated. An adaptive inverter comprising at least a quantization width determining circuit that determines a quantization width, and a multiplier that generates an output signal by multiplying the current output signal of the quantization width determining circuit by the output of the inverse quantization circuit. An adaptive quantization and inverse quantization device comprising a quantization circuit.