JP3010655B2 - Compression encoding apparatus and method, and decoding apparatus and method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a compression encoding apparatus and method for performing predictive analysis of an input digital signal, and a decoding apparatus and method.

[Summary of the Invention]

The present invention provides a compression encoding apparatus that obtains a difference between a prediction value signal from a prediction filter and an input digital signal, and that quantizes and transmits parameters of a prediction filter. Compression encoding apparatus and method capable of performing encoding with high quantization efficiency by quantizing parameters using a quantization table selected according to the properties of an input digital signal. , And a decoding device and method.

[Conventional technology]

In general, for an audio signal, a video signal, or the like, information can be compressed by encoding a difference signal (prediction residual) between a value predicted using a correlation or the like between adjacent samples and an actual sample value. . As a prediction method of such predictive coding, for example, linear predictive analysis (linear predictive codin
g: LPC) etc. are applied. Here, the linear prediction analysis is a method of adjusting a prediction filter coefficient as a parameter representing a linear prediction filter to minimize the sum of squares of the difference signal (prediction residual). At this time, the coefficient of the prediction filter is a ladder filter coefficient or a lattice filter coefficient, or a line spectrum pair (LSP).
r) coefficients and their modified filter coefficients.

For example, according to the analysis of a speech waveform to which the method of linear prediction analysis is applied, the properties of the speech waveform and its spectrum can be efficiently and accurately expressed with a very small number of parameters.
Moreover, those parameters can be obtained by relatively simple calculations.

Conventionally, such parameters have been quantized to low bits by quantization using a predetermined quantization table. In this case, a specific example of the quantization using the quantization table includes, for example, vector quantization using a single codebook. This vector quantization refers to a comparison between an input vector composed of a plurality of the above parameters and each code vector in a single codebook (codebook) composed of a plurality of code vectors (representative vectors) ( For example, comparison by distance calculation or the like) is performed, and an optimum code vector most similar (closest to the distance) to the input vector is selected, and an identification code (index) corresponding to the optimum code vector is used as a quantization output. What you get.

Here, FIG. 7 shows a configuration of a partial autocorrelation type speech analysis / synthesis system of an input speech signal (that is, a configuration of encoding and decoding). In the configuration of FIG. 7, each filter stage is configured in a lattice form, and the lattice filter coefficients (k of the lattice filter
) Is quantized by a predetermined quantization table.

In FIG. 7, the analog audio signal waveform supplied via the input terminal 101 of the audio analysis system (encoding) is
An analog / digital (A / D) converter performs sampling at a predetermined sampling frequency, and this waveform data is added to adders 103 _{1 to} 103 _p , 108 _{1 to} 108 _p , and multipliers 104 _{1 to} 104
_p , 106 _{1 to} 106 _p , delay device 107 _{1 to} 107 _p , correlation calculator 105 ₁ to 10
In 5 _p it is transmitted to the configured lattice prediction filter circuit 115. Waveform data supplied to the lattice-shaped prediction filter circuit 115, together with the first sent to the correlation calculator 105 _1, delayer
Also sent to 107 _1, after being delayed a predetermined time in the delay unit 107 _1, are sent to the correlation computing unit 105 _1. From correlation calculator 105 _1, so that the above k parameter (k ₁₎ is obtained by the correlation calculation between the waveform data and the delayed data. The k parameter (k ₁ ) is calculated by the multipliers 104 ₁ and 106 ₁
Sent to The multiplier 104 the waveform data is _1, the aforementioned multiplier 106 ₁ is the delayed data is supplied, the multipliers 104 _1, 106 ₁ respectively supplied data to the k parameter (k ₁₎ Is multiplied. The output from the multiplier 104 ₁ is transmitted to the adder 108 ₁ by being subtracted from the delayed data, the difference data is obtained. Also a multiplier
The output from 106 ₁ may be transmitted to the adder 103 ₁ by being subtracted from the waveform data, the difference data is obtained. Hereinafter, adders 103 _{2 to} 103 _p , 108 _{2 to} 108 _p , multipliers 104 _{2 to} 104 _p ,
106 _{2 to} 106 _p , delay device 107 _{2 to} 107 _p , correlation calculator 105 _{2 to} 105 _p
Performs a similar filtering process.

By such filtering processing is performed, the adder 103 the prediction residual waveform data, which is the difference _p input signal and the prediction value signal is obtained, the prediction residual waveform data, transmitted to the correlation detection circuit 109 Is done. In the correlation detection circuit 109, amplitude information A and pitch information T are obtained from the predicted residual waveform data.

Each of the k parameters (k _{1 to} k _p ) from each of the correlation calculators 105 _{1 to} 105 _p is transmitted to the encoder 110 and is quantized by a predetermined quantization table, that is, for example, the single code described above. Output after being subjected to quantization processing by vector quantization using a book. The amplitude information A
And pitch information T are also output via the encoder 110. These outputs are encoded outputs.

Next, as a speech synthesis system (decoding), the encoder 110
Are sent to the decoder 120, and the encoder 110
, The inverse quantization (inverse vector quantization) processing of the vector quantization processing is performed. The pitch information T via the decoder 120 is transmitted to the changeover switch 1 via the pulse generator 121.
Transmitted to 23. Also, random noise from the random noise generator 122 is sent to the changeover switch 123. The changeover switch 123 is switched between voiced sound and unvoiced sound, and the output is sent to the multiplier 124. Further, the multiplier 124 has the amplitude information A
, And data approximating the above-mentioned predicted residual waveform data can be obtained from the multiplier 124.

The estimated residual waveform approximation data is added to adders 125 _{1 to} 125 _p , 127
_{1 to} 127 _p , multipliers 126 _{1 to} 126 _p , 128 _{1 to} 128 _p , delay unit 129 ₁ to
It is supplied to the constructed lattice prediction filter 135 in 129 _p. Each of the multipliers 126 _{1 to} 126 _p and 128 _{1 to} 128 _p receives each of the decoded (inverse vector quantized) k parameters. In the lattice prediction filter circuit 135, the prediction residual waveform approximation data is filtered through a path opposite to that of the lattice prediction filter 115. However, the delay units 129 _{1 to} 129 _p and the adders 127 _{1 to} 12
7 _p, the data to be sequentially processed by the multiplier 128 ₁ to 128 _p is an output of the adder 125 ₁ after passing through the respective adders 125 ₁ to 125 _p.

Data approximating the waveform data obtained by such a filtering process is converted into an analog synthesized voice waveform by a digital / analog (D / A) converter 130 and output from an output terminal 141.

[Problems to be solved by the invention]

Incidentally, audio signals, video signals, and the like can be divided into several groups depending on the properties of the signals. For example, an audio signal is composed of a voiced sound and an unvoiced sound having different properties from each other, and can be divided into such a voiced sound group and an unvoiced sound group.
The voiced sound is a general term for a sound accompanied by vocal cord vibration, and the unvoiced sound is a generic term for a sound not accompanied by vocal cord vibration.

In such a case, the characteristics of the linear prediction filter for signals grouped according to their properties are different from each other. That is, since the distribution of the correlation of the parameters (distribution of the pattern) is different,
In the case where those parameters are vector-quantized by a predetermined quantization table, for example, by a single codebook,
The signal compression efficiency (quantization efficiency) decreases. In other words,
The transmission bit rate will drop.

Therefore, the present invention has been proposed in view of the above situation, and performs encoding of an input digital signal with high quantization efficiency even for an input digital signal composed of signals having different properties. It is an object of the present invention to provide a compression encoding apparatus and method, and a decoding apparatus and method capable of performing the above.

[Means for solving the problem]

A compression encoding apparatus and method according to the present invention have been proposed in order to solve the above-described problem, and determine a difference between a predicted value signal from a prediction filter that predicts and analyzes an input digital signal and the input digital signal. And a plurality of types of quantization tables provided for quantizing the parameters according to the properties of signals based on voiced and unvoiced sounds of the input signal in the compression encoding for quantizing and transmitting the parameters of the prediction filter. And quantizing the parameter based on the controlled quantization table.

Further, the decoding apparatus and method according to the present invention provide an input signal based on a difference between a predicted value signal from a prediction filter for predictively analyzing a digital signal and the input digital signal, and a parameter signal obtained by quantizing parameters of the prediction filter. When a digital signal is restored from the input signal and the parameter signal, a plurality of types of signals provided for dequantizing the parameter are provided depending on the nature of the signal based on the voiced sound and the unvoiced sound of the input signal. The quantization table is controlled, and the parameter is inversely quantized based on the controlled quantization table.

[Action]

According to the present invention, a property of an input digital signal is determined,
Since the parameters are quantized using the quantization table selected according to the property, the number of bits allocated at the time of parameter quantization can be reduced.

〔Example〕

Hereinafter, embodiments to which the compression encoding apparatus and method and the decoding apparatus and method according to the present invention are applied will be described with reference to the drawings.

FIG. 1 shows a schematic configuration of a compression encoding apparatus according to an embodiment of the present invention. The present embodiment apparatus obtains a difference (prediction residual) between a prediction value signal from a prediction filter that predicts and analyzes an input digital signal (for example, a voice digital signal or the like) and the input digital signal, and sets parameters of the prediction filter. In a compression encoding apparatus for quantizing and transmitting, a plurality of types of quantization tables for quantizing the above parameters are provided in advance in accordance with the characteristics of an input signal, for example, voiced sound and unvoiced sound in an audio signal. The plurality of types of quantization tables are stored in the quantization table storage memory 7, for example. The quantization of the parameters is performed by a quantization table selected from a plurality of types of quantization tables in the quantization table storage memory 7 according to the properties of the input digital signal. The properties of the input digital signal at this time are required. That is, for example, the correlation detection / determination circuit 6 determines the property of the voice signal as a voiced sound or an unvoiced sound.

In FIG. 1, for example, a digital audio signal is supplied as an input digital signal to an input terminal 1 of the compression encoding apparatus of the present embodiment, and this audio signal is first supplied to an adder 2 and a prediction filter 3. Transmitted. At the same time, the audio signal is also transmitted to the linear prediction analysis circuit 4. In the linear prediction analysis circuit 4, the audio signal is framed at predetermined time intervals, and linear prediction analysis is performed for each frame to determine the parameters of the prediction filter 3. According to the parameters obtained in this way,
In the prediction filter 3, the audio signal is filtered, and the output of the prediction value signal obtained from the prediction filter 3 is transmitted to the adder 2. The adder 2
, A difference (prediction residual) between the audio signal and the predicted value signal is obtained. The parameters are output from an output terminal 11 after being quantized by a parameter quantization circuit 8 that quantizes the parameters (for example, vector quantization).

Here, the parameter quantization in the parameter quantization circuit 8 is performed based on a plurality of types of quantization tables prepared (stored) in the quantization table storage memory 7 based on the properties of the voice signal, that is, voiced and unvoiced sounds. This is performed by a quantization table selected according to the property. That is, in the parameter quantization circuit 8, for example, vector quantization of the above parameters is performed.
At the time of the vector quantization, a codebook corresponding to the voiced sound or the unvoiced sound is selected from a plurality of codebooks (quantization tables) stored in the quantization table storage memory 7, and the selection is performed. A code vector in the code book thus obtained is compared with an input vector composed of a plurality of parameters. An identification code by this vector quantization is obtained as a quantized output of the parameter quantization circuit 8.

The prediction residual is transmitted to the correlation detection / determination circuit 6 in order to determine whether it is a voiced sound or an unvoiced sound as described above. The correlation detection / discrimination circuit 6 discriminates between voiced sound and unvoiced sound of the audio signal. That is, in general, the voiced sound signal of the audio signal is a signal having a waveform having a periodic peak as shown in FIG. 2, for example, and the unvoiced sound signal is a signal having a waveform as shown in FIG. 3, for example. The voiced sound and the unvoiced sound are basically different sounds. Here, as an example of the method of property discrimination performed by the correlation detection / discrimination circuit 6, for example, property discrimination based on a normalized autocorrelation coefficient of the prediction residual can be mentioned. That is, the normalized autocorrelation coefficient of the prediction residual in the frame unit is obtained, and according to the maximum value of the normalized autocorrelation coefficient,
A determination is made of the nature of the input digital signal. For example, when the input digital signal is a voice digital signal, it is determined whether the input voice digital signal is a voiced signal or an unvoiced sound signal. At this time,
The normalized autocorrelation coefficient r (i) of the prediction residual is Can be represented by Note that when −1 ≦ r (i) ≦ 1 and N
Is the number of samples. Here, the frame is 20 msec,
The normalized autocorrelation coefficient r (i) of the audio digital signal having a sample frequency of 8 kHz and N = 160 is, for example, the fourth
FIG. 5 and FIG. FIG. 4 shows the normalized autocorrelation coefficient r for voiced sounds and FIG. 5 shows the normalized autocorrelation coefficient r for unvoiced sounds.
The example of (i) is shown. 4 and 5 that the voiced sound and the unvoiced sound are completely different sounds. That is, the correlation detection / determination circuit 6 determines that the sound is a voiced sound if the maximum value of the normalized autocorrelation coefficient r (i) obtained as described above is larger than a predetermined value. If it is smaller than the predetermined value, it is determined that the sound is unvoiced.

Further, in the correlation detection / discrimination circuit 6, the amplitude information A and the pitch information T are obtained from the prediction residual, similarly to the correlation detection circuit 109 in FIG. Therefore, the correlation detection
For example, the discrimination circuit 6 can also perform a property discrimination such that the pitch information T is “0” and the voice information is unvoiced, and when the pitch information T is not “0”, the voice data is a voiced sound. The amplitude information A is output from the output terminal 9 and the pitch information T is output from the output terminal 10.

For example, the property discrimination information obtained as described above is transmitted to the quantization table storage memory 7, and as described above, the quantization table storage memory 7 outputs either the voiced sound or the unvoiced sound. The selected quantization table (codebook) is selected, and vector quantization is performed based on the selected quantization table.

Further, as parameters from the linear prediction analysis circuit 4 in FIG. 1, ladder filter coefficients, lattice filter coefficients, line spectrum pair coefficients (LSP coefficients), and their modified filter coefficients can be used. Also,
The correlation detection / discrimination circuit 6 may be configured not to obtain the amplitude information A and the pitch information T from the prediction residual as described above, but to output the prediction residual as it is. . In this case, the decoding apparatus is configured to obtain the property discrimination information from the prediction residual.

When the partial autocorrelation coefficient is used as the parameter, the adder 2, the prediction filter 3,
Prediction filter circuit section 5 composed of linear prediction analysis circuit 4
Can be realized by a configuration like the lattice-type prediction filter circuit 115 in FIG. In this case, the linear prediction analysis circuit 4 corresponds to the correlation calculators 105 _{1 to} 105 _p in FIG. 7, and the adder 2 and the prediction filter 3 are the adders 103 _{1 to} 103 _p and 108 _{1 in} FIG. -108 _p, the multiplier 104 ₁ ~104 _p, 106 ₁ ~10
6 _p corresponds to the delay units 107 _{1 to} 107 _p . Further, the correlation detection / determination circuit 6 corresponds to the correlation detection circuit 109, and the encoder 110 performs vector quantization according to the properties of the signal.

 FIG. 6 shows a schematic configuration of the decoding device row.

In FIG. 6, the amplitude information A is input to the input terminal 21 of the decoding device, the pitch information T is input to the input terminal 22,
The input terminal 23 is supplied with the quantized parameters.

First, the quantized parameter is calculated by an inverse quantization circuit.
Transmitted to 33. Further, the pitch information T is transmitted to a voiced / unvoiced detection circuit 35 which detects whether the pitch information T is a voiced sound or an unvoiced sound. In the voiced / unvoiced detection circuit 35, from the pitch information T, information for determining the nature of voiced sound or unvoiced sound is extracted. The property discrimination information from the voiced / unvoiced detection circuit 35 is transmitted to the quantization table storage memory 32 which stores the same quantization table as the quantization table storage memory 7 of the encoding apparatus in FIG. . From the quantization table storage memory 32, data of a quantization table based on the voiced sound or the unvoiced sound is output and transmitted to the inverse quantization circuit 33. The inverse quantization circuit 33 includes a first
It performs the reverse process of the parameter quantization circuit 8 in the figure. Therefore, the inverse quantization circuit 33 performs inverse vector quantization of the quantized parameters using the quantization table (code book) from the quantization table storage memory 32.

The amplitude information A is sent to the pulse generation circuit 25. The pitch information T is also supplied to the pulse generation circuit 25. Therefore, the pulse generation circuit 25
A pulse signal based on the amplitude information A and the pitch information T is output. This pulse signal is transmitted to the changeover switch 27 which is switched and controlled based on the voiced / unvoiced sound discrimination information from the voiced / unvoiced detection circuit 35. The random noise from the random noise generator 26 is sent to the changeover switch 27. Therefore, the changeover switch 26 outputs each pulse of the pulse signal and the random noise in accordance with the voiced sound or the unvoiced sound. Are switched and output. The output of the changeover switch 26 is transmitted to an adder 29 via a gain control amplifier 28 controlled based on the amplitude information A. The output from the gain control amplifier 28 is data that approximates the prediction residual.

The parameters from the inverse quantization circuit 33 are transmitted to the prediction filter 30. This prediction filter 30
The output of the adder 29 is subjected to filtering processing by the parameter from the inverse quantization circuit 33, and the prediction value signal from the prediction filter 30 is fed back to the adder 29, so that the output from the adder 29 is The addition processing of the prediction residual approximation data and the prediction signal is performed to obtain a decoded digital signal. This decoded digital signal is output from the output terminal 34.

When the partial autocorrelation coefficient is used as the parameter, the adder 29 surrounded by a broken line in FIG.
The prediction filter circuit unit 31 composed of 30
It can be realized by a configuration like the lattice type prediction filter circuit 135 in the figure. In this case, the adder 29 and the prediction filter 30 are composed of the adders 125 _{1 to} 125 _p , 127 _{1 to} 127 _p , the multipliers 126 _{1 to} 264 _p , 128 _{1 to} 128 _p , and the delay units 129 ₁ to 127 in FIG. This corresponds to 129 _p .

In the present embodiment, as described above, by performing parameter quantization using a quantization table corresponding to the nature of a voiced or unvoiced signal, it is possible to reduce the number of bits allocated for the prediction filter as a result. Thus, distortion of frequency characteristics due to quantization can be reduced. Further, from the viewpoint of the entire compression encoding apparatus, a high-quality signal can be restored at a low bit rate.

The input digital signal described above is not limited to the digital audio signal, but may be a general audio signal including a musical instrument sound or the like, or may be a video digital signal. Further, the nature of the audio signal is not limited to voiced and unvoiced sounds.

〔The invention's effect〕

ADVANTAGE OF THE INVENTION According to the compression encoding apparatus and method of this invention, while obtaining the difference between the said input digital signal and the prediction value signal from the prediction filter which predicts and analyzes an input digital signal, the parameter of the said prediction filter is quantized and transmitted. At that time,
A plurality of types of quantization tables provided for quantizing the parameters are controlled according to the properties of signals based on voiced and unvoiced sounds of the input signal, and the parameters are quantized based on the controlled quantization tables. By doing so, it is possible to reduce the amount of computation for searching for the optimum output value at the time of quantization, shorten the encoding processing time, and perform the same for input signals having different characteristics between voiced and unvoiced sounds. Encoding can be performed with high quantization efficiency. That is, from the viewpoint of the entire compression encoding apparatus, the number of bits allocated for parameter quantization can be reduced, and furthermore, distortion of frequency characteristics due to parameter quantization can be reduced.

Therefore, when decoding, a high-quality signal can be restored at a low bit rate in the entire apparatus.

[Brief description of the drawings]

1 is a block circuit diagram showing a schematic configuration of a compression encoding apparatus according to an embodiment of the present invention, FIG. 2 is a waveform diagram of a voiced signal, FIG. 3 is a waveform diagram of an unvoiced signal, and FIG. FIG. 5 is a characteristic diagram showing a normalized autocorrelation coefficient of a vocal signal, FIG. 5 is a characteristic diagram showing a normalized autocorrelation coefficient of an unvoiced signal, and FIG. 6 shows a schematic configuration of a decoding device. FIG. 7 is a block circuit diagram showing the configuration of a conventional partial autocorrelation type speech analysis / synthesis system. 2 ... Adder 3 ... Prediction filter 4 ... Linear prediction analysis circuit 5 ... Prediction filter circuit section 6 ... Correlation detection / discrimination circuit 7 ... Quantization table storage memory 8 ... Parameter quantization circuit

Claims (4)

(57) [Claims] 1. A compression encoding apparatus for obtaining a difference between a predicted value signal from a prediction filter for predicting and analyzing an input digital signal and the input digital signal, and quantizing and transmitting parameters of the prediction filter. A voiced sound, by a property of a signal based on an unvoiced sound, a control unit that controls a plurality of types of quantization tables provided for quantizing the parameter, based on the controlled quantization table, And a quantizing unit for quantizing. 2. A compression encoding method for obtaining a difference between a predicted value signal from a prediction filter for predicting and analyzing an input digital signal and the input digital signal, and quantizing and transmitting parameters of the prediction filter. Controls a plurality of types of quantization tables provided for quantizing the parameters according to the properties of signals based on voiced and unvoiced sounds, and quantizes the parameters based on the controlled quantization tables. A compression encoding method characterized by the above-mentioned. 3. An input signal based on a difference between a prediction value signal from a prediction filter for predicting and analyzing a digital signal and the input digital signal, and a parameter signal obtained by quantizing parameters of the prediction filter. In a decoding device for restoring a digital signal from an input signal and a parameter signal, a plurality of types of quantization tables provided for dequantizing the parameter are controlled according to the properties of a signal based on voiced and unvoiced sounds of the input signal. A decoding device comprising: a control unit; and an inverse quantization unit that inversely quantizes the parameter based on the controlled quantization table. 4. An input signal based on a difference between a predicted value signal from a prediction filter for predicting and analyzing a digital signal and the input digital signal, and a parameter signal obtained by quantizing parameters of the prediction filter. In a decoding method for restoring a digital signal from an input signal and a parameter signal, a plurality of types of quantization tables provided for dequantizing the parameter are controlled according to the properties of a signal based on voiced and unvoiced sounds of the input signal. A decoding method for inversely quantizing the parameter based on the controlled quantization table.