JP2507431B2 - ADPCM encoding / decoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial field of application] The present invention relates to adaptive differential PCM (hereinafter referred to as ADPCM) encoding
The present invention relates to a decoding method, and more particularly, to an ADPCM coding / decoding method which discriminates between a modem signal and a voice signal with high accuracy and has a good coding / decoding characteristic of the modem signal in the voice signal band.

[Prior Art] Regarding the conventional ADPCM encoding / decoding method,
EE, Telecommunications Conference, 1984, 23.1.1. ~ 23.1.4 (IEEE, TELECOMMUNICAT
IONS CONFERENCE, 1984, 23.1.1-23.1.4). The method of identifying the input signal in this conventional technique will be described below. The conventional method has a quantization width YU _j that performs a high-speed operation that can sufficiently cope with a voice signal with a large amplitude change in the input signal, and a quantization width YL _j that performs a low-speed operation that can adapt to a modem signal with a small amplitude change. _j is separately generated, and a linear combination of these two quantization widths is used to generate a true quantization width _Yj for actual quantization. That is, Y _j = A _j · YU _j-1 + (1-A _j ) YL _{j-1 is} calculated using A _j as a linear combination constant to obtain the quantization width Y _j . (here,
j indicates the sampling time at which processing is performed, and the “quantization width” corresponds to the “scale factor” in the paper. )
The constant A _j is a short time average (averaging operation with a long time constant) of a weighting function F (I _j ) having a quantized value I _j obtained as a result of quantization with a quantization width Y _j. The time average is obtained, and it is obtained by determining whether the input signal is a voice signal or a modem signal based on the difference between them. That is, D _{ms, j} = (1-2 ^-5 ) D _{ms, j-1} +2 ^-5 F (I _j ) D _{ml, j} = (1-2 ^-7 ) D _{ml, j-1} +2 ^-7 F (I _j ) to obtain a long-time average D _{ml, j} and a short-time average D _{ms, j} , and A _{p, j} = (1-2 ^-4 ) A _{p, j-1} +2 ^-3 (Δ _j > 2 ^-3 D _{ml, j} ) = (1-2 ^-4 ) A _{p, j-1} (when Δ _j ≤2 ^-3 D _{ml, j} ) However, the constant A _j is obtained by Δ _j = | D _{ms, j} −D _{ml, j} |. The meaning of this equation is that if the amplitude change of the input signal is small, the long-term average D _{ml , j}
, The difference Δ _j should be small, and conversely, if Δ _j is large, it means that the statistical properties of the input signal are not equal and the change is rapid like a voice signal. It is reasonable to estimate.

[Problems to be Solved by the Invention] Consider a case where a modem signal in the voice signal band, for example, a 9.6 kbps modem signal is input to the ADPCM encoder. A signal having a constant statistical property, such as a modem signal, also has a small signal correlation, and thus is very difficult to predict. It is therefore the input to the quantizer of the ADPCM encoder. The difference signal D _j between the input signal and the prediction signal is not very small. Therefore, the quantized value I _j also changes considerably. For example, for a phase-modulated modem signal, the difference signal D _j at a large phase change is quite large. Therefore, the short-time average D _ms also changes considerably, and Δ _j does not always decrease. Therefore, there is a problem in that the input signal cannot be identified properly, the coding distortion of the modem signal increases, and a bit error occurs.

An object of the present invention is to provide an ADPCM coding / decoding method having good coding characteristics not only for speech but also for signals in other speech bands.

[Means for Solving Problems] The above-described object is to perform adaptive pole prediction for an input signal having a large correlation such as voice, and fixed pole prediction for a signal having a small correlation such as a modem signal. As a prediction signal, the linear combination constant is controlled by whether or not the prediction coefficient of the adaptive polar prediction is within a predetermined variation region, and the quantum of the quantizer is controlled. As for the width of the signal, a signal that has a distribution characteristic in which the amplitude is several times as large as the standard deviation value and the amplitude appears, such as a voice signal, is adapted to change rapidly and greatly. For the stationary signal of, the quantization width is adapted to change slowly and in small steps so as not to change the quantization width unnecessarily, and the linear combination of the quantization widths generated by these two adaptations is taken. Quantization width, linear at this time The coupling constant is also achieved by controlling the prediction coefficient of the adaptive polar prediction in accordance with whether or not the prediction coefficient of the adaptive polar prediction is within a predetermined region, like the linear combination constant in the prediction.

[Operation] For a signal that has a spectrum bandwidth that occupies almost the entire voice band, such as a 9.6 kbps modem signal, and that has little correlation with the signal, the prediction gain is obtained by performing fixed pole prediction that matches the spectrum. It can be earned, and the adaptation method of the quantization width matches the characteristics of the modem signal by performing the adaptation that has a slow and small response after convergence to the quantization width that matches the input signal level. ADPCM encoding / decoding can be performed. In ADPCM encoding / decoding for signals other than modem signals, such as voice signals, for polar prediction output and quantization width output, the linear combination of the voice signal and the modem signal is taken as the true polar prediction output, The true quantization width output is used, and both of these two sets of linear combination constants are used for the modem polar prediction side depending on whether or not the value of the prediction coefficient of the adaptive polar prediction is within a predetermined fluctuation range for the modem signal. And, it controls to the adaptive quantization width side for the modem or another set of the polar prediction side for the voice and the quantization width side, so that the encoding / decoding can be performed according to the property of the input signal.
Coding distortion can be reduced.

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

1 and 2 are ADs according to an embodiment of the present invention, respectively.
ADPCM encoder and ADPC to which PCM encoding / decoding method is applied
It is a block diagram of an M decoder. Further, FIG. 3 is a prediction coefficient determination area diagram used when determining whether or not the input signal is a modem signal.

In FIG. 1, 1 is an input terminal, 2 is a difference circuit, 3 is a quantizer, 4 is an inverse quantizer, 5 is a fast adaptive quantization width generator, 6 is a slow adaptive quantization width generator, and 7 is Linear combination constant generator, 8 is an adaptive zero predictor, 9 is an adaptive pole predictor, 10 is a fixed pole predictor, 11 is a modem signal discriminator, 12 is a linear combination constant generator, 13 to 16 are multipliers, 17 21 is an adder, and 22 is an output terminal.

Now, assuming that the signal S _j is input to the input terminal 1 at time j, the difference between the signal S _j and the prediction signal SE _j output from the adder 19 is taken by the difference circuit 2, and this difference signal D _j is obtained. The ADPCM code value I _j quantized by the quantizer 3 is output from the output terminal 22.
At the same time, the signal I _j is input to the quantization width generators 5 and 6 that perform an adaptive operation at high speed or low speed, and the quantization width is updated. For example, citing the description of the conventional example, the output of the fast adaptive quantization width generator 5 is YU _j , and the output of the slow adaptive quantization width generator 6 is YL _j . Linear coupling constant generator 7
Is output by the modem signal discriminator 11, which will be described in detail later.
A linear coupling constant β _j is generated, and for example, Is calculated. Using this output β _{j + 1} , the true quantization width Y _{j + 1} is output from the adder 21.

Y _{j + 1} = β _{j + 1} · YU _{j + 1} + (1-β _{j + 1} ) YL _{j + 1} On the other hand, the ADPCM code value I _j is dequantized by the dequantizer 4 and the dequantized value DQ _j Is output. Based on this, the adaptive zero predictor 8 performs adaptive zero prediction, and the zero predicted value SEZ _{j + 1} is output. For example, Is calculated.

The adder 17 outputs the polar prediction residual, and outputs DQ _j + SEZ _j .

The adder 18 outputs the reproduction signal SR _j, and outputs a value of SR _j = SP _j + DQ _j + SEZ _j = SE _j + DQ _j . Where SP _j is the pole predictor and the adder 20
Is the output of.

Based on this reproduced signal SR _j , the adaptive pole predictor 9 performs adaptive pole prediction and the fixed pole predictor 10 performs fixed pole prediction, respectively, and the pole prediction values are obtained. Is output. For example, Is calculated.

Is a fixed prediction constant, and the frequency spectrum of the modem signal is determined by the transfer function of the fixed pole predictor 10. Should be chosen so that

The modem signal discriminator 11 determines whether or not the prediction coefficients ▲ a ¹ _j ▼, ▲ a ² _j ▼ updated at every sample according to the above equation fall within the area A as shown in FIG. Then, it is determined that the signal is a modem signal when entering the area A, and is a signal other than the modem signal when outside the area A. Then, based on the result, the linear coupling constant generator 12 generates the linear coupling constant α _j . For example, Is calculated.

This coupling constant α _j is multiplied by the output of the adaptive pole predictor 9 by the multiplier 15, 1-α _j is multiplied by the output of the fixed pole predictor 10 by the multiplier 16, and the outputs of both multipliers 15, 16 are added. By adding at 20, the pole prediction value SP _j is obtained.

The adder 19 obtains the prediction signal SE _j from the zero prediction value SEZ _j and the pole prediction value SP _j, and is calculated by SE _j = SP _j + SEZ _j .

The adder 21 generates a quantization width Y _{j + 1} following the input signal level. Based on this, the difference signal D _j from the difference circuit 2 is added to the ADPCM code value I _{j by the} quantizer 3. On the other hand, the inverse quantizer 4 also quantizes the ADPCM code value I from the quantizer 3 as
_By dequantizing _j as the dequantized value DQ _{j with the} quantization width Y _{j + 1} , the original difference signal D _j is restored.

In the ADPCM decoder of FIG. 2, 23 is an input terminal, 24 is an inverse quantizer, 25 is a fast adaptive quantization width generator, 26 is a slow adaptive quantization width generator, 27 is a linear combination constant generator, 28 Is an adaptive zero predictor, 29 is an adaptive pole predictor, 30 is a fixed pole predictor, 31 is a modem signal discriminator, 32 is a linear combination constant generator, 33 to 36 are multipliers, 37 to 40 are adders, 41 Is an output terminal.

The operation on the decoding side is exactly the same as the description after outputting the ADPCM code value I _{j in} the description on the code side, and outputs the reproduction signal SR _j to the output terminal 4
Output to 1.

FIG. 3 is a judgment area diagram for judging whether or not it is a modem signal by the polar prediction coefficients ▲ a ¹ _j ▼ and ▲ a ² _j ▼, as described above. When it is outside A, it is determined to be a signal other than the modem. As the area A, for example, You can choose like. In fact, the polar prediction coefficient is rje
^If we have ^-jωj as a pole, Where r represents the bandwidth information and ω represents the center frequency information of the spectrum. And
For modem signals, spectrum bandwidth Sadamari by the modulation scheme, the spectrum density substantially uniform distribution in the range r is not much change, since ω gather near the carrier frequency of the modem signal, ▲ a ¹ The variation range of _j ▼, ▲ a ² _j ▼ can be limited to a certain area (area A in the example of FIG. 3). On the other hand, in the case of an audio signal, most of the spectrum concentrates on the low speed, and therefore concentrates on the parts other than the area A in the first and fourth quadrants of FIG.

In the case of tone signals, the spectrum band is narrow, so
Concentrate on the area of a ₂ <−0.73, that is, the base of the triangle.

As described above, it is determined whether or not ▲ a ¹ _j ▼ and ▲ a ² _j ▼ are included in the area A of FIG. 3, the result is filtered, and the smoothed one is used as the linear combination constant. As a result, stable switching is performed, and the above determination can be made accurately because the prediction coefficients are concentrated in a region determined by the frequency characteristics of the input signal, particularly in the case of a modem signal, the bandwidth and carrier frequency.

[Effect of the Invention] According to the present invention, it is possible to accurately determine whether or not the input signal is a modem signal, and thus it is possible to realize a suitable ADPCM encoding / decoding method with no encoding distortion for the modem signal.

[Brief description of drawings]

1 and 2 show ADPCM according to an embodiment of the present invention.
FIG. 3 is a block configuration diagram of an encoder and a decoder to which the encoding / decoding method is applied, and FIG. 3 is a determination region diagram for determining whether or not an input signal is a modem signal by a polar prediction coefficient. 3 ... Quantizer, 4, 24 ... Inverse quantizer, 5, 25 ... High-speed adaptive quantization width generator, 6, 26 ... Low-speed adaptive quantization width generator, 7, 12, 27,
32 ... Linear combination constant generator, 8, 28 ... Adaptive zero predictor, 9, 29 ...
Adaptive pole predictor, 10,30 ... Fixed pole predictor, 11,31 ... Modem signal discriminator.

Claims (1)

(57) [Claims] 1. An ADPCM encoding / decoding method for quantizing and encoding / decoding a difference signal obtained by subtracting a prediction signal of the input signal from a sampled input signal of a digital code. The quantization width to be used is generated by a linear combination of the fast adaptive quantization width and the slow adaptive quantization width generated by the two adaptation methods of high speed and low speed, and the polar prediction of the prediction signal is quantized. Generated by the linear combination of the adaptively predicted value and the fixedly predicted value from the reproduced signal obtained by adding the difference signal and the predicted signal, and used for the linear combination and the linear combination of the quantization widths. The ADPCM code characterized in that the linear combination constant values to be determined are determined by whether or not the two types of adaptive pole prediction coefficient values reflecting the frequency characteristics of the input signal are within a predetermined fluctuation region. Encryption / decryption method.