JP3211762B2 - Audio and music coding - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech encoding system, and more particularly to a system for encoding both speech and music with high efficiency and high quality at a minimum necessary bit rate.

[0002]

2. Description of the Related Art As a code-excited linear predictive speech coding apparatus for obtaining a natural decoded sound even when background noise or music is added to speech, for example, Japanese Patent Application Laid-Open No. 8-123493 discloses a codebook (codebook). Consists of multiple subcodebooks,
Each sub-codebook is composed of a speech part consisting of a signal source vector learned using speech without noise and a signal source vector learned using non-speech other than speech or a random signal source vector not learning. And a codebook search is performed, thereby improving the voice quality.

[0003]

However, the conventional code-driven linear predictive speech coding system has the following problems.

[0004] The first problem is that when music is encoded and reproduced using the conventional code-driven linear predictive speech encoding method, the sound quality is greatly deteriorated.

[0005] The reason is that the conventional code-driven linear predictive speech coding method is a linear predictive speech coding method for speech based on a speech synthesis model.

The second problem is that the conventional music encoding system has good quality when encoding and reproducing audio, but has a lower encoding efficiency than the audio encoding system.

[0007] The reason is that the bit rate becomes higher than necessary compared to the speech coding method.

SUMMARY OF THE INVENTION Accordingly, the present invention has been made in view of the above-described problems, and has as its object to provide a high-efficiency and high-quality code for both speech and music at the minimum necessary bit rate. To provide a voice and music encoding system.

[0009]

In order to achieve the above object, a speech and music encoding system according to the present invention performs code-driven linear predictive speech encoding of an input signal, and an error between the input speech and reproduced speech is determined in advance. While the code-driven linear predictive speech coding is continuously selected while the difference between the input signal and the reproduced voice is larger than a predetermined threshold while the code-driven linear predictive speech coding is performed. If it is determined that the encoding cannot be performed with sufficiently high quality by the method, a means for selecting music encoding from the next frame is provided. Hereinafter, the present invention will be described in accordance with preferred embodiments.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS In a preferred embodiment of the present invention, a linear prediction analysis means (1 in FIG. 1) for linearly analyzing an input speech signal to obtain a synthesis filter coefficient, and It is obtained by driving a means for calculating information (2 in FIG. 1), a pitch reproduction filter composed of the pitch information, and a cascade filter of the synthesis filter with a code included in a codebook (4 in FIG. 1). Codebook searching means (3 in FIG. 1) for searching the codebook and determining a code so that an error signal of the input voice signal is minimized; Means for encoding a code number corresponding to the determined code (5 in FIG. 1)
, A code-driven linear predictive speech encoding means comprising: a means for encoding an input signal into music (7 in FIG. 1); and a means for detecting the power of the input signal (8 in FIG. 1). In the code-driven linear predictive speech encoding means performs encoding, while the error signal does not exceed the first threshold, encoding by the code-driven linear predictive speech encoding means,
When the error signal exceeds the first threshold, the input audio signal is switched to perform music encoding, and when the power of the input audio signal becomes equal to or less than the second threshold. Switching to perform encoding with the code-driven linear predictive speech encoding means, and a selection signal indicating which of the code-driven linear predictive speech encoding and the music encoding has been selected, and a selected code. (6 in FIG. 1) for multiplexing and outputting.

On the decoding side, when the selection signal indicating which of the code-driven linear predictive speech coding and the music coding is selected indicates that the code-driven linear predictive speech coding is selected, Means for decoding the encoded synthesis filter coefficient, pitch information and code number inputted from the encoding side (21 in FIG. 2), and a pitch reproduction filter (FIG. 2) composed of the decoded pitch information. 2 of 2
3) and a cascade filter (24 in FIG. 2) of a synthesis filter composed of the decoded synthesis filter coefficients, and a code book is obtained from the decoded code number by referring to a code book. Means for reproducing sound by driving the cascade filter (22 in FIG. 2);
When the selection signal indicates that music encoding is selected, a means for decoding music (25 in FIG. 2) and a code-driven linear predictive speech encoding or music encoding which receives the selection signal and receives the selection signal. Means (26 in FIG. 1) for switching and outputting the reproduced sound or the music decoded output from the cascade filter according to the selection.

In the system disclosed in Japanese Patent Application Laid-Open No. 8-123493, a code book is composed of a code book for a voice part and a code book for a non-voice part with respect to a code-driven linear predictive voice coding method. A search is performed to improve voice quality.

On the other hand, according to the present invention, when it is determined that the code-driven linear predictive speech coding method can perform coding with sufficiently high quality, the code-driven linear predictive speech coding method is selected. If it is determined that the encoding cannot be performed with sufficiently high quality by the linear predictive speech coding method, by selecting a music coding method, for both voice and music, each with a minimum necessary bit rate, Encode with high efficiency and high quality.

[0014]

Next, in order to describe the above-mentioned embodiment of the present invention in more detail, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of an audio / music encoding encoder according to the present invention.

Referring to FIG. 1, an encoder input terminal 10
Is supplied to the linear prediction analyzer 1, the pitch extractor 2, the codebook searcher 3, and the power detector 8.

The linear prediction analyzer 1 linearly analyzes the speech signal X (n) to obtain a coefficient α of a linear prediction filter, and supplies the coefficient α to the codebook searcher 3 and the encoder 5.

The pitch extractor 2 uses the audio signal X (n)
The pitch information including the pitch period and the pitch gain is extracted and supplied to the codebook searcher 3 and the encoder 5.

The codebook searcher 3 forms a pitch reproduction filter from the pitch information, forms a linear prediction filter from the coefficient α of the linear prediction filter, and cascades the pitch reproduction filter and the linear prediction filter into a codebook 4.
The code book is searched and the code is determined so that the error E between the synthesized voice and the original voice driven by the code included in the code book n (1 ≦ n ≦ N) prepared in the above is minimized. Is supplied to the encoder 5. The error E between the synthesized speech and the original speech is supplied to the switch 6.

The code book 4 has the same code book on the encoding side and the decoding side.

The encoder 5 encodes the linear prediction filter coefficient α, the pitch information and the code number, and supplies them to the switch 6.

The above processing is performed by code-driven linear prediction (CEL).
P: Code Excited Linear Prediction).

The music encoder 7 performs music encoding on the audio signal X (n), and switches the obtained code to the switching unit 6.
To supply.

The power detector 8 has an encoder input terminal 10
The power of the audio signal X (n) supplied from is calculated and supplied to the switch 6.

The switch 6 is based on the error E between the synthesized speech and the original speech supplied from the codebook searcher 3 and the power of the speech signal X (n) supplied from the power detector 8.
It controls switching between code-driven linear predictive speech coding and music coding.

Hereinafter, the operation of the switch 6 will be described.

First, the switch 6 selects code-driven linear prediction speech coding and performs processing.

While the error E between the synthesized speech and the original speech supplied from the codebook searcher 3 does not exceed the threshold value A, the switch 6 selects the code-driven linear predictive speech coding and continues the processing. .

On the other hand, if the error E exceeds the threshold value A, the switch 6 switches to music coding and performs processing.

When the power of the audio signal X (n) supplied from the power detector 8 falls below the threshold value B, the switch 6 switches to code-driven linear predictive audio coding again.
Perform processing.

The switch 6 multiplexes a selection signal indicating which of the code-driven linear predictive speech coding and the music coding has been selected and the selected code, and outputs the signal to the encoder output terminal 11.
To supply.

Here, the selected codes are coded linear prediction filter coefficients α, pitch information and code numbers in the case of code-driven linear prediction speech coding, and in the case of music coding, It is a speech code.

As music encoding, MPEG2
Music coding such as Audio may be used, or
If not so high quality is required, audio encoding such as ADPCM may be performed without much deterioration.

FIG. 2 is a block diagram showing the configuration of an embodiment of a speech / music encoding decoder according to the present invention.

Referring to FIG. 2, the decoder input terminal 30
, A selection signal indicating which of the code-driven linear predictive speech coding and the music coding has been selected and the selected code are input from the coding side.

Here, the selected codes are coded linear prediction filter coefficients α, pitch information, and code numbers in the case of code-driven linear prediction speech coding, and in the case of music coding, , A speech code.

The coded linear prediction filter coefficient α, pitch information and code number supplied to the decoder input terminal 30 are supplied to the decoder 21.

The music code supplied to the decoder input terminal 30 is supplied to the music decoder 25.

The selection signal, which is supplied to the decoder input terminal 30 and indicates which of the code-driven linear predictive speech coding and the music coding has been selected, is supplied to the switch 26.

The switch 26 performs code-driven linear predictive speech decoding when the selection signal indicating which of code-driven linear predictive speech coding and music coding is selected indicates code-driven linear predictive speech coding. . That is, the following processing is performed.

The coded linear prediction filter coefficient α, pitch information and code number supplied to the decoder 21 are decoded, and the decoded linear prediction filter coefficient α is decoded by the synthesis filter 24. The pitch information is supplied to a pitch reproduction filter 23, and the decoded code number is supplied to a drive signal generator 22.

The drive signal synthesizer 22 includes a code book 27
The first drive signal is synthesized by reading out the code corresponding to the code number from the code book n (1 ≦ n ≦ N) stored in the first drive signal and supplied to the pitch reproduction filter 23.

The pitch reproduction filter 23 forms a pitch reproduction filter based on the input pitch information, and drives the pitch reproduction filter with the first drive signal.
The second drive signal is synthesized and supplied to the linear prediction filter 25.

The linear prediction filter 24 forms a linear prediction filter from the input linear prediction filter coefficient α, and generates a synthesized speech by driving the linear prediction filter with the second drive signal. The output is supplied to a decoder output terminal 31.

The switch 26 performs music decoding when the selection signal indicating which of the code-driven linear predictive speech coding and the music coding is selected indicates the music coding.
That is, the following processing is performed.

The music code supplied to the music decoder 25 is decoded by the music decoder 25 to reproduce a music signal, supplied to the switch 26, and output to the decoder output terminal 31.
Supplied to

[0047]

As described above, according to the present invention, the following effects can be obtained.

A first effect of the present invention is that, when speech is input, a code-driven linear predictive speech encoding method which requires only a minimum bit rate necessary for encoding speech is automatically selected. In addition, the encoding can be performed with high efficiency and high quality.

The reason is that, in the present invention, when speech is input and it is determined that high-quality coding can be performed by the code-driven linear prediction speech coding method, the code-driven linear prediction speech coding method is automatically selected. This is because the configuration is such that

A second effect of the present invention is that, when music is input, the most suitable music coding method for coding music can be automatically selected, and high efficiency and high quality coding can be performed. It is.

The reason is that when music is input and it is determined that high-quality coding cannot be performed by the code-driven linear predictive speech coding method, the music coding method is automatically selected.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of an embodiment of an audio / music encoding encoder according to the present invention.

FIG. 2 is a block diagram showing a configuration of an embodiment of a speech / music encoding decoder according to the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Linear prediction analyzer 2 Pitch information calculator 3 Codebook searcher 4 Codebook 5 Encoder 6 Switching device 7 Music encoder 8 Power detector 10 Encoder input terminal 11 Encoder output terminal 21 Decoder Reference Signs List 22 drive signal synthesizer 23 pitch reproduction filter 24 linear prediction filter 25 music decoder 26 switcher 27 codebook 30 decoder input terminal 31 decoder output terminal

Claims (3)

(57) [Claims] A linear prediction analysis means for performing linear prediction analysis of an input audio signal to obtain a synthesis filter coefficient; a means for calculating pitch information from the input audio signal; a pitch reproduction filter comprising the pitch information; A synthesized speech obtained by driving a cascade filter of a synthesis filter with a code included in a code book, and a code book that searches the code book and determines a code so that an error signal of the input speech signal is minimized. Searching means; coding means for coding a code number corresponding to the synthesized filter coefficient, the pitch information, and the determined code; and a code-driven linear predictive voice coding means comprising: Means for detecting the power of the input audio signal; and Means for encoding, while the error signal does not exceed the first threshold, perform the encoding in the code-driven linear predictive speech encoding means, the error signal exceeds the first threshold In this case, the input audio signal is switched to perform music encoding, and when the power of the input audio signal becomes equal to or less than a second threshold, the code is again encoded by the code-driven linear predictive audio encoding means. And switch to
Means for multiplexing and outputting a selection signal indicating which of the code-driven linear predictive speech coding and the music coding has been selected, and a selected code; apparatus. 2. The audio and music encoding apparatus according to claim 1,
A selection signal that is output from the device and indicates whether code-driven linear prediction speech coding or music coding has been selected, and a selection signal .
To the audio and music decoding device that inputs the selected code.
There are, the selection signal, to indicate that the code excited linear predictive speech coding is selected, according to claim 1 before, wherein
Means for decoding encoded synthesis filter coefficients (linear prediction coefficients), pitch information, and code numbers input from the speech and music encoding device; and a pitch reproduction filter composed of the decoded pitch information. And a cascade filter comprising a synthesis filter composed of the decoded synthesis filter coefficients; and obtaining a code by referring to a codebook from the decoded code number, and driving the cascade filter with the obtained code. 2. The audio and music encoding device according to claim 1, wherein the audio and music encoding device according to claim 1, wherein the audio signal is reproduced, and when the selection signal indicates that music encoding is selected.
The cascade depending on whether the music decryption hand <br/> stage for music decoding music codes are more input, the inputs of the selection signal code excited linear predictive speech coding or music coding is selected off the output of the output or the music decryption means from the filter
Speech and music decoding apparatus characterized by comprising: means for changing output, a. 3. A coding apparatus comprising: a linear prediction analysis means for performing linear prediction analysis on an input speech signal to obtain a synthesis filter coefficient; a means for calculating pitch information from the input speech signal; Searching the code book and synthesizing the code so that the synthesized speech obtained by driving the cascade filter of the reproduction filter and the synthesis filter with the code included in the code book and the error signal of the input speech signal are minimized. A code book search means for determining; and a means for encoding a code number corresponding to the synthesized filter coefficient, the pitch information, and the determined code. Means for music encoding the audio signal; and means for detecting the power of the input audio signal. Encoding is performed by the speech measurement encoding means, and while the error signal does not exceed the first threshold, encoding is performed by the code-driven linear prediction audio encoding means, and the error signal is encoded by the first signal. If the threshold value is exceeded, the input audio signal is switched to perform music encoding, and if the power of the input audio signal falls below a second threshold, the code-driven linear predictive audio encoding is performed. Means for switching to perform encoding by means, and a means for multiplexing and outputting a selection signal indicating which of the code-driven linear predictive speech encoding and the music encoding has been selected, and the selected code. The decoding side has a selection signal indicating which of the code-driven linear predictive speech coding and the music coding has been selected,
Means for decoding the coded synthesis filter coefficient, pitch information, and code number input from the coding side when the code-driven linear prediction speech coding is selected, A cascade filter composed of a pitch reproduction filter composed of the decoded pitch information and a synthesis filter composed of the decoded synthesis filter coefficients; and obtaining a code by referring to a codebook from the decoded code number. Means for reproducing sound by driving the cascade filter with the determined code; means for decoding music when the selection signal indicates that music encoding is selected; Inputting a signal and decoding the reproduced sound or music from the cascade filter depending on whether code-driven linear predictive speech coding or music coding is selected. Means for outputting an output.