JP3205161B2 - Audio coding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice coder used for digital telephones and digital recorders which require high efficiency voice compression.

[0002]

2. Description of the Related Art In recent years, since a medium for transmitting or storing data is limited, a speech encoding apparatus which realizes a high speech quality and a high compression rate has been desired. In particular, it is important to introduce an adaptive codebook capable of adaptively reflecting the characteristics of the input speech in order to achieve high efficiency.

[0003] An example of a conventional analysis-synthesis type speech coding apparatus will be described below. FIG. 4 shows a block diagram of a conventional speech coding apparatus. In FIG. 4, reference numeral 41 denotes a voice input device for inputting voice. Reference numeral 42 denotes a fundamental frequency extractor, which extracts a fundamental frequency of the input voice. Reference numeral 43 denotes a frequency amplitude calculator that calculates an amplitude value of a harmonic frequency according to the extracted fundamental frequency.

[0004] In the speech coding apparatus configured as described above, the fundamental frequency extractor 42 extracts the power spectrum and the fundamental frequency shown in FIG. 6 from the speech having the waveform shown in FIG. Is extracted. Thereafter, the amplitude of the harmonic frequency shown in FIG. 7 is obtained by the frequency amplitude calculator 42 based on the fundamental frequency.

FIG. 8 shows a synthesized speech waveform.

[0006]

However, in the above-mentioned conventional analysis-synthesis type speech coding apparatus, since the phase information of the input speech is not used when analyzing the input speech, the input speech waveform shown in FIG. 8, the synthesized speech waveforms are completely different and cannot be directly compared. Therefore, the parameters obtained by the analysis for each frame must be quantized by a fixed codebook prepared in advance, and the quantization efficiency is low.

The present invention solves the above-mentioned conventional problem, and provides an excellent speech coding apparatus that performs efficient quantization by introducing an adaptive codebook that adaptively changes while reflecting the features of speech. The purpose is to do.

[0008]

In order to achieve the above object, the present invention provides a voice input device for inputting voice, an input voice autocorrelation function calculator for calculating an autocorrelation function of the input voice, and an input device. A fundamental frequency extractor for extracting a fundamental frequency of speech, a speech analyzer for analyzing an input speech based on the extracted fundamental frequency and a vector parameter obtained by a parameter reader and extracting parameters, and A speech synthesizer that generates a synthesized speech according to the generated parameters, a synthesized speech autocorrelation function calculator that obtains an autocorrelation function of the generated synthesized speech, and an adaptive codebook having the synthesized speech autocorrelation function as a code vector. A parameter calculator for extracting a change rate of a parameter at a point where an autocorrelation function is obtained, and a code vector for calculating the change rate parameter. A parameter codebook having the same as the input code, an adaptive codebook searcher for obtaining a code number of a function closest to the autocorrelation function of the input voice from the adaptive codebook, and a vector parameter indicated by the obtained code number. And a parameter reader for reading from the parameter code book.

[0009]

With this configuration, an adaptive codebook composed of the autocorrelation function of the synthesized speech is searched by the autocorrelation function of the input speech, and the input speech is converted using the change rate parameter of the code number obtained by the search. Because of the analysis, the input speech can be directly compared with the synthesized speech. That is, it is possible to use an adaptive codebook that changes following the characteristics of the input voice, and the parameters can be efficiently quantized according to the characteristics of the input voice.

[0010]

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

In FIG. 1, reference numeral 11 denotes a voice input unit for inputting voice, and 12 denotes a fundamental frequency extractor, which extracts a fundamental frequency of the input voice. Reference numeral 13 denotes a voice analyzer that analyzes an input voice based on the extracted fundamental frequency. Reference numeral 14 denotes an input speech autocorrelation function calculator, which calculates an autocorrelation function of the input speech. An adaptive codebook search unit 15 selects a code vector close to the autocorrelation function from an adaptive codebook 20 described later. Reference numeral 16 denotes a parameter reader which reads out the obtained code number change rate parameter from the parameter code book 17. This parameter codebook 17 is composed of a change rate parameter calculated by a parameter calculator 18 for a change rate of a synthesis parameter used for speech synthesis. Reference numeral 19 denotes a speech synthesizer, which generates a synthesized speech according to the parameters obtained by the speech analyzer 13. The adaptive codebook 20 stores the autocorrelation function of the speech generated by the speech synthesizer 19 calculated by the synthesized speech autocorrelation function calculator 21.

Next, the operation of the speech coding apparatus configured as described above will be described with reference to FIGS.

First, an autocorrelation function of the speech input by the speech input device 11 is calculated by an input speech autocorrelation function calculator 14.
Is calculated by The adaptive codebook searcher 15 searches the adaptive codebook 20 for a code vector close to the autocorrelation function. The adaptive code book 20 includes an autocorrelation function of the synthesized speech generated by the speech synthesizer 19 shown in FIG.
The one shown in FIG.

On the other hand, the fundamental frequency of the input speech is extracted by the fundamental frequency analyzer 12, and the input speech is analyzed by the speech analyzer 13 based on the extracted fundamental frequency. Here, when analyzing the input speech, the change rate parameter of the code number selected by the adaptive codebook searcher 15 is read from the parameter codebook 17 by the parameter reader 16, and the parameter and the fundamental frequency obtained earlier are used. Analyze the input speech.

Next, the operation of the voice analyzer 13 will be described. The parameter code book 17 is composed of the rate-of-change parameters obtained by the parameter calculator 18, and as shown in FIG. 3, the parameter change in a certain section L for obtaining the auto-correlation function calculated by the auto-correlation function calculator 21. Represents the rate. Therefore, the parameters extracted by the parameter reader 16 indicate the rate of change of the parameters in the analysis frame.

In the voice analyzer 13, the fundamental frequency analyzer 1
The voice is analyzed on the basis of the line spectrum when the frequency is changed in accordance with the rate of change of the above-mentioned parameter, centering on the fundamental frequency obtained in Step 2.

As is apparent from the above, in the speech coding apparatus according to the present embodiment, an adaptive codebook method is introduced, and a change rate parameter obtained by the adaptive codebook method is used.
An excellent effect is obtained in that efficient quantization can be performed by a code book reflecting the characteristics of speech, and speech can be dynamically analyzed.

As described above, according to the present embodiment, an adaptive codebook having an autocorrelation function as a codebook and a parameter codebook corresponding to the adaptive codebook are quantized and synthesized by a codebook containing speech characteristics. Since the analysis can be performed in consideration of the change in the parameter, the analysis and encoding can be performed efficiently and dynamically.

[0019]

As described above, the present invention provides an autocorrelation function calculator for a synthesized speech, an adaptive codebook using the autocorrelation function as a code vector, and a parameter codebook using a speech synthesis parameter as a code vector. This enables dynamic speech analysis according to the characteristics of the input speech, and realizes an excellent high-efficiency speech encoding device.

[Brief description of the drawings]

FIG. 1 is a block diagram of a speech encoding apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a relationship between a synthesized speech and an adaptive codebook in the embodiment.

FIG. 3 is an explanatory diagram showing a relationship between a synthesized speech and a parameter code book in the embodiment.

FIG. 4 is a block diagram of a conventional speech encoding device.

FIG. 5 is a conventional input voice waveform diagram.

FIG. 6 is an explanatory diagram showing a power spectrum and a fundamental frequency of an input voice in the related art.

FIG. 7 is an explanatory diagram showing an amplitude value of a harmonic frequency in the related art.

FIG. 8 is a conventional synthetic speech waveform diagram.

[Explanation of symbols]

 Reference Signs List 11 voice input device 12 fundamental frequency extractor 13 voice analyzer 14 input voice autocorrelation function calculator 15 adaptive codebook searcher 16 parameter reader 17 parameter codebook 18 parameter calculator 19 voice synthesizer 20 adaptive codebook 21 synthesized voice Autocorrelation function calculator 41 voice input device 42 fundamental frequency extractor 43 frequency amplitude calculator

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁷ , DB name) G10L 19/00-19/14 H03M 7/30 H04B 14/04

Claims (1)

(57) [Claims] A voice input device for inputting voice; an input voice autocorrelation function calculator for calculating an autocorrelation function of the input voice; a fundamental frequency extractor for extracting a fundamental frequency of the input voice; A voice analyzer that analyzes the input voice based on the obtained fundamental frequency and the vector parameter obtained by the parameter reader and extracts the parameter;
A speech synthesizer for generating a synthesized speech according to the extracted parameters, a synthesized speech autocorrelation function calculator for obtaining an autocorrelation function of the generated synthesized speech, and an adaptive code having the synthesized speech autocorrelation function as a code vector Book, a parameter calculator for extracting a change rate of a parameter at a point where an autocorrelation function is obtained, a parameter code book having the change rate parameter as a code vector, and a code of a function closest to the autocorrelation function of the input voice. An audio coding apparatus comprising: an adaptive codebook searcher for obtaining a number from the adaptive codebook; and a parameter reader for reading a vector parameter indicated by the obtained code number from the parameter codebook.