JPS62194743A - Adpcm coding and decoding device - Google Patents

Abstract

PURPOSE:To improve the forecast characteristic by constituting a forecast means by a broad band fixed pole forecast device and an adaptive zero forecast device so as to improve the frequency characteristic of the pole forecast signal with respect to a MODEM signal. CONSTITUTION:A residual signal 9 being the subtraction of a zero forecast signal 7 and a broad band fixed pole forecast signal 8 from an input signal 6 sampled at the coder side is inputted to an adaptive quantizer 1. A signal 10 quantized by the quantizer 1 is sent to the decoder side. On the other hand, the sent quantizing signal 10 at the decoder side is subject to inverse quantization by an adaptive inverse quantizer 2 and a residual signal 12 is recovered. Then a zero forecast signal 13 and a broad band fixed pole forecast signal 14 are added to the residual signal 12 to recover a recovery signal 15. Thus, the ADPCM coder and decoder is realized, where an excellent transmission characteristic is obtained for the MODEM signal having 9,600bps or over of transmission speed.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to an adaptive differential PCM (ADPCM) encoding method that quantizes and transmits only the difference between an input signal and its predicted value, and is particularly applicable to audio signals and ADPCM code, which is effective for compressing and transmitting voice band signals such as data modem signals.
Regarding the decoder.

[Conventional technology]

Conventionally, input signal prediction in an ADPCM coder/decoder that compresses and transmits voice band signals has been described, for example, in IE, Global Telecommunications Conference (1984), pp. 774-7.
Page 77 (IEEE, Global Talt
comb-nication conference
e (1984) PP 774 777), a prediction method was used in which a pole predictor and a zero predictor are adaptively changed according to the input signal.

A block diagram of a conventional ADPCM encoder and decoder is shown in FIG. In FIG. 8, 1 is an adaptive quantizer, 2 is an adaptive inverse quantizer, 3 is an adaptive zero predictor, 5 is a zero predictor adaptive control unit, 24 is an adaptive pole predictor, and 25 is a pole predictor adaptive control unit. part, 6 is the input signal, 26.27 is the polar prediction signal,
8 and 13 are zero prediction signals, 11° and 15 are reproduced signals, 9.
12 is a residual signal, and 10 is a quantized signal.

The operation of the conventional ADPCM encoder and decoder will be described below with reference to FIG. On the encoder side, it is 8kHz/A residual signal 9 obtained by subtracting the zero prediction signal 8 and the pole prediction signal 26 from the blinged input signal 6 is input to the adaptive quantizer 1. Then, 4-bit quantization is performed by the adaptive quantizer 1, and the quantized signal 10 is transmitted to the decoder side. On the post-decoder side, the transmitted quantized signal 1o is inversely quantized by the adaptive inverse quantizer 2, and the residual signal 1o is
Play 2. Then, the zero prediction signal 13 and the polar prediction signal 27 are added to the reproduced residual signal 12 to obtain the reproduction signal 1.
Play 5.

Here, the adaptive zero predictor 3 inputs the inverse quantized signal from the inverse quantizer 2, obtains amplitude information of the input signal 6 from this inverse quantized signal, predicts the next input signal, and generates a zero prediction signal 8. or 13
occurs. The frequency characteristics of the adaptive zero predictor 3 are adaptively controlled by the zero predictor adaptive control section 5 so as to follow the frequency of the input signal.

Similarly, the adaptive pole predictor 24 adds the zero prediction signal and the pole prediction signal to the dequantized signal and reproduces the reproduced signal 11 or 1.
5 is input, the next input signal is predicted based on this reproduced signal, and a polar prediction signal 26 or 27 is generated. The frequency characteristics of the adaptive pole predictor 24 are adaptively controlled by the pole predictor adaptive control unit 26 so as to follow the frequency of the input signal.

As described above, according to the conventional technology, the residual signal 9 becomes smaller than the input signal 6 due to the effects of the adaptive zero predictor 3 and the adaptive pole predictor 24. less than when quantizing, resulting in 8 k
A Hz x 4 bit = 32 kAps K signal can be compressed (64 kbpz if uncompressed).

[Problem that the invention seeks to solve]

In the above conventional technology, audio signals and 4800 hpz 'j
However, it was difficult to transmit modem signals of 9600 bpz or higher, which have recently become a growing demand.

Hereinafter, using FIG. 9, in the conventional technology, 960°hp
The reason why it is difficult to transmit modem signals of s or more will be explained.

FIG. 9 is a diagram explaining the frequency response characteristic of the polar prediction signal 26 in the prior art, in which 20 is 9600 bpz.
Average frequency characteristics of the modem signal, 28 is an example of the frequency characteristics of a 9600 bpr modem signal considered in a short period of time within a few samples (SiCω) (hereinafter abbreviated as short-time frequency characteristics of the modem signal), 29 is an adaptive pole predictor 24 is the frequency characteristic H'Cω), and 30 is the frequency characteristic Sε'(ω) of the pole prediction signal 26 for the short-time modem signal.

As shown in Fig. 9, the average frequency characteristic of the 9600 bpz modem signal is a broadband characteristic of the band 500#z to 2900Hz, but the short-time frequency characteristic of the modem signal Si
(ω) is a narrow band characteristic having a peak within the band (500 Hz to 2900 #z), and the position of the peak moves with time. On the other hand, the frequency characteristic H'(ω) of the conventional adaptive pole predictor 24 moves adaptively according to the short-term frequency characteristic target (ω)K of the modem signal, but it is almost at the center of the band. It had a narrow band characteristic with a peak around 1700Hz. This is because the control of frequency characteristics cannot keep up with modem signals of 9600 pbs or more, and cannot follow the short-term frequency characteristics si(ω) of the modem signals.

Now, as shown in Fig. 9, the short-time frequency characteristic Si of the modem signal is
Consider the case where the peak position of (ω) is far from 1700 Hz. Frequency characteristic Sg'(
ω) is given by the product of the frequency characteristic H'(ω) of the adaptive pole predictor 240 and the frequency characteristic sr(ω) of the reproduced signal.

st' (ω) = H' (ω) x Sr (ω) (1) where,
Since Sr(ω) is approximately equal to the short-time frequency characteristic Si(ω) of the modem signal, which is the input signal, equation (1) can be expressed as Sg'(
ω) #H'(ω) x si (ω) (2). From equation (2), the frequency characteristic 54' of the polar prediction signal 26 (
ω) has a characteristic in which the peak portion of the short-time frequency characteristic si(ω) of the modem signal, which is the input signal, is almost cut off, as shown in FIG. to degrade. As a result, the prediction characteristics of the predictor deteriorate, and the transmission characteristics of the 9600 bps modem deteriorate.

In this way, with the conventional technology, the frequency characteristic of a 9600 bps modem signal considered within a single time period of several samples is 17
When the peak is located away from 00H2, the frequency response characteristic of the polar prediction signal to the modem signal is deteriorated, the prediction performance is deteriorated, and there is a problem that transmission by a 9600 hpz modem becomes difficult.

An object of the present invention is to improve the frequency response characteristic of a polar prediction signal to a modem signal, thereby improving the prediction characteristic.
The object of the present invention is to provide an ADPCM encoder/decoder that can obtain good transmission characteristics even for modem signals of bpz or higher.

[Means for solving problems]

The purpose of the above is to configure the pole predictor not with an adaptive pole predictor whose frequency characteristics change depending on the frequency of the input signal, but with a wideband fixed pole predictor that satisfies the frequency characteristics of the input signal. However, this is achieved by predicting the input signal by combining this broadband fixed pole predictor and an adaptive zero predictor for improving amplitude change tracking characteristics.

[Effect]

In the present invention, the fixed pole predictor has a wide band fixed frequency characteristic that covers the frequency of the input signal, and the frequency characteristic does not change depending on the frequency of the input signal as in the conventional case.

Therefore, even if the speed of the input signal increases, the frequency characteristics of the input signal will not be fully followed and the response characteristics will not deteriorate, and the predictive characteristics will not deteriorate.

In addition, if the number of delay stages is increased in order to make the frequency characteristics of the fixed pole predictor broadband, there is a risk that it will not be able to follow sudden changes in the amplitude of the input signal. Since the amplitude tracking characteristic is adaptively dealt with, the amplitude tracking characteristic is improved.

〔Example〕

An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an ADPCM encoder and decoder according to an embodiment of the present invention, in which 1 is an adaptive quantizer, 2 is an adaptive inverse quantizer, 3 is an adaptive zero predictor, and 4 is a wideband fixed pole. Predictor, 5 is a zero predictor adaptive control unit 6 is an input signal, 7.14 is a broadband fixed pole prediction signal, 8.13 is a zero prediction signal, 9,1
2 is a residual signal, 11.15 is a reproduced signal, and 10 is a quantized signal.

In FIG. 1, on the encoder side, a residual signal 9 obtained by subtracting a zero prediction signal 7 and a wideband fixed pole prediction signal 8 from an input signal 6 which is sampled at BkHz is input to an adaptive quantizer 1, and is subjected to 4-bit quantization. The quantized signal 10 is then transmitted to the decoder side. On the other hand, on the decoder side, the transmitted quantized signal 10 is inversely quantized by an adaptive inverse quantizer 2, and a residual signal 12 is reproduced.

Then, the zero prediction signal 13 is added to the reproduced residual signal 12.
and the wideband fixed pole prediction signal 14 are added to reproduce the reproduced signal 15. In the above operation, the difference from the conventional example shown in FIG.
The prediction characteristics of the broadband fixed pole predictor 4) for a 9600 hpz modem signal.

FIG. 2 is a block diagram showing an example of the adaptive zero predictor 3, in which 16 is a delay device and 17 is a variable multiplier. In the variable multiplier 17, the multiplication coefficients h1 to h are varied by the zero predictor adaptive control section 5 according to the frequency of the input signal. In FIG. 2, six stages of the delay device 16 are provided, but the difference in stage of the delay device is not limited to this and may be arbitrary.

Further, FIG. 3 is a block diagram showing an example of the wideband fixed pole predictor 4, in which 18 is a delay device and 19 is a fixed multiplier. In the wideband fixed pole predictor, each multiplication coefficient of the fixed multiplier 19 is a fixed value, and the frequency characteristics are determined by the value of each multiplication coefficient and the step of the delay device. For example, a fixed multiplier with 10
FIG. 4 shows an outline of the frequency characteristics when stages are provided and each multiplication coefficient is set to the following values.

α, = 1.064476, α, = −2
,004551α,=2.027038,α
, = -2681159α, = 2.209509
, α, = −2,300597α7→1.
470075, α, = −1,266507
α, = 0.5483817, α1゜= -0,
In the case of 380988, a frequency characteristic that satisfies a signal in the audio band can be obtained.

Next, the frequency response characteristics of the broadband fixed pole prediction signal 7 in this case will be explained using FIG. 5.

In FIG. 5, 20 is the average frequency characteristic of the 9600 Apt modem signal, 21 is the short-time frequency characteristic of this modem signal (SiCω), 22 is the wideband fixed pole predictor 40 frequency characteristic H(ω), and 26 is the short-time frequency characteristic of the modem signal. This is the frequency characteristic Sg(, ω) of the broadband fixed pole prediction signal 7 with respect to the modem signal. At this time, as in equation (2) above, the frequency characteristic Sg(ω) of the broadband fixed pole prediction signal 7 is given by the product of Si(ω) and H(ω), so 5g(ω) is shown in FIG. As shown, the characteristic has a peak at the same location as the short-time frequency characteristic Si(ω) of the modem signal, and the frequency response characteristic is improved.

This is true regardless of where the peak of the short-time frequency characteristic Si(ω) of the modem signal is within the band of the modem signal.

Incidentally, the frequency characteristic of the fixed pole predictor is expressed by the following equation, where the pole prediction signal is 1(A), the reproduction signal fsrck) is the prediction coefficient, and α is the prediction coefficient.

5sCk)=Σ α6Sr (ki) --
(311; However, N is the number of stages of the delay device of the predictor.

Therefore, in order to make the fixed pole predictor broadband as in the present invention, it is necessary to increase the number of stages N.

However, if the number of stages N is increased, for example, if the amplitude of the input signal suddenly changes from dog to small, the negative effect of the past reproduced signal 5r(k i) with a large amplitude will continue for a long time, and the amplitude of the polar prediction signal will decrease. The sharp bends may not be reduced, and the amplitude change tracking characteristics may deteriorate.

In this example, for such amplitude change tracking characteristics,
The adaptive zero predictor 3 works effectively. The adaptive zero predictor 3 has almost no frequency characteristics and adaptively responds to the difference between the input signal and the fixed pole prediction signal, so it functions to compensate for and improve the deterioration of the amplitude change tracking characteristic of the fixed pole prediction signal. do.

The operation of such an adaptive zero predictor will be explained using FIGS. 6 and 7.

FIG. 6 is a diagram showing an example of a prediction signal for a 9600 bpz modem signal, where (α) is for the conventional example shown in FIG. 8, and +A) is for the case of the conventional example shown in FIG. In this case, (c) is the case where the broadband fixed pole predictor 4 and the adaptive zero predictor 3 of this embodiment are used. In addition, Figure 7 shows the case where the amplitude of the input signal changes suddenly at 960O
These are diagrams showing actual examples of prediction signals for AP & modem signals, where (α) is the case where only the wideband fixed pole predictor 4 is used as the predictor in this embodiment, and (h) is the case where the predictor is the wideband fixed pole predictor 4 of this embodiment. and the case where the adaptive zero predictor 3 is used. In each figure, the solid line indicates a 9600 phz modem signal, which is an input signal, and the broken line indicates a predicted signal. Note that in FIGS. 6 and 7, the predicted signal is the sum of the zero predicted signal 8 and the polar predicted signal 26 in FIG. 8 in the conventional example, and the zero predicted signal 8 in FIG. 1 in the present embodiment. and the sum of the wideband fixed-pole prediction signal 7, and the wideband fixed-pole prediction signal 7 when only the wideband fixed-pole predictor is used as the predictor in this embodiment.

First, the frequency response of the predicted signal to the 9600 APJI modem signal will be explained using FIG.

In the conventional example, as shown in FIG. 3 (α), the frequency response of the predicted signal was poor, and the predicted signal could not follow the phase change of the input signal (see the arrow). In this embodiment, when the predictor is only the broadband fixed pole predictor 4, the predicted signal can better follow the phase change of the input signal than in the conventional example, as shown in FIG. 3(a), and the frequency response characteristic is improved. In this embodiment, when the adaptive zero predictor 3 is also used, the third
As shown in Figure (c), the characteristics of the predicted signal are almost the same as in the case of only the broadband fixed predictor, and in this case, the adaptive zero predictor 3 is not significantly involved in improving the frequency response characteristics. However, the adaptive zero predictor 6 has an effect of improving the amplitude change tracking characteristic, which is a weak point of the wideband fixed pole predictor.

The effect of the adaptive zero predictor 3 will be explained using FIG. 4. In this embodiment, when the predictor is only the broadband fixed pole predictor 4, as shown in FIG. 4 (α), when the amplitude of the input signal changes suddenly from large to small, the amplitude change of the predicted signal is The characteristics are poor, and the predicted signal does not become small, causing a phenomenon similar to oscillation (see the arrow). In contrast, when the adaptive zero predictor 3 is also used in this embodiment, the amplitude change tracking characteristic is improved by the effect of the adaptive zero predictor 6, as shown in FIG. 4 (side view).

As described above, according to this embodiment, it is possible to improve the frequency response characteristics and amplitude change tracking characteristics of a predicted signal for a 9600 hpz modem signal. As a result, the prediction gain is increased and the maximum prediction residual value is reduced compared to the conventional method, and the prediction performance is improved, so that the transmission characteristics of the ADPCM encoder/decoder for 9600 hpt modem signals can be improved.

The table below shows the prediction gain and maximum prediction residual value when the number of stages of the delay device and each multiplication coefficient of the wideband fixed pole predictor are set to the values described above. It can be arbitrarily determined depending on the frequency characteristics.

However, S-th = input signal d: prediction residual signal σ, two-person signal effective value In addition, in this embodiment, since the polar predictor is fixed, the amount of processing is less than that of the conventional method, and the audio signal and 96001r
It is possible to realize an ARPCM encoder/decoder that is compatible with both pz modem signals.

〔Effect of the invention〕

According to the present invention, it is possible to improve the prediction performance of the predictor in the ADPCM encoder/decoder.
It is possible to realize an ADPCM encoder/decoder having good transmission characteristics even for modem signals of bpz or higher.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an ADPCM encoder and decoder according to an embodiment of the present invention, FIG. 2 is a block diagram of an adaptive zero predictor in FIG. 1, and FIG. 3 is a block diagram of an adaptive zero predictor in FIG. 1. 4 is a schematic diagram of the frequency characteristics of the wideband fixed pole predictor shown in FIG. 3, and FIG. 5 is a diagram showing the frequency response characteristics of the wideband fixed pole prediction signal. Figure 7 is 9600p
A diagram showing an example of a predicted signal for a modem signal of bz,
FIG. 8 is a block diagram of a conventional ADPCM encoder and decoder, and FIG. 9 is a diagram showing frequency response characteristics of a conventional adaptive pole prediction signal. 1......Adaptive quantizer 2...Adaptive inverse quantizer 3.
......Adaptive zero predictor 4...
......Broadband fixed pole predictor 5...
......Zero predictor adaptive control section 6...
...... Input signal 7.14 ... Wideband fixed pole prediction signal 8.13 ... Zero prediction signal 9.12 ... Residual signal 10... Quantized signal 11, 15
...Playback signal agent Patent attorney Katsuo Ogawa 11 Kuchi A () PCM' (@Mokikan 2 Adai 313S] 1 only 34 Kuchi 16 figure] Ura v8 Zu ＾+J
A 8th A() PCI'lτ Kamo Rashi 19 Figure

Claims (1)

[Claims] 1. A means for predicting the value of the next input signal from the value of the input signal is provided, and the residual signal obtained by subtracting the predicted signal obtained from the above prediction means from the input signal is quantum-coupled with a step size corresponding to the magnitude of the amplitude. quantization/inverse quantization, transmission/reproduction A
In the DPCM code/decoder, the prediction means includes a fixed pole predictor having a wideband frequency characteristic corresponding to the frequency characteristic of the input signal, and an adaptive zero predictor whose frequency characteristic changes according to the frequency of the input signal. An ADPCM encoder/decoder comprising: