JPH0782358B2 - Speech analysis and synthesis method - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech analysis / synthesis method for reducing the amount of information used in speech synthesis without degrading the quality of synthesized speech in a speech analysis / synthesis system.

Configuration of Conventional Example and Its Problems In general, a speech analysis and synthesis method is to analyze a speech signal at a frame period m, as shown in FIG. 1, and determine a section type (silent section, unvoiced section, voiced section), spectrum parameter. , A sound source amplitude control parameter and a driving sound source period parameter are extracted, and a voice is synthesized from these parameters and the driving sound source by using a digital filter. Each parameter is updated every frame period m.

FIG. 2 shows a sound source amplitude control parameter B, a driving sound source cycle parameter C, and an original sound waveform A obtained by analyzing the sound. As is clear from FIG. 2, the driving sound source cycle parameter (FIG. 2C) often has the same value in adjacent frames. Therefore, when the driving sound source periods of adjacent frames have the same value, the repeat flag can be set to achieve information compression. With this method, about 60% of the driving sound source information is reduced. However, as for the sound source amplitude control parameter (FIG. 2B), since the values of adjacent frames are almost the same, information compression is not performed.

FIG. 3 shows a frame which must have a sound source amplitude control parameter value or a driving sound source cycle parameter value when speech is synthesized by a conventional synthesis method. The drive sound source cycle parameter is only 40% of all frames have a value due to the repeat flag.
The sound source amplitude control parameter has its value in every frame. As described above, the conventional speech analysis and synthesis method in which the information is compressed by setting the repeat flag only when the parameter values of the adjacent frames are the same has a drawback that the information of the sound source amplitude control parameter cannot be compressed.

An object of the present invention is to eliminate the above-mentioned drawbacks of the conventional example, and to provide a voice analysis / synthesis method for performing information compression of a sound source amplitude control parameter and a driving sound source period parameter without deteriorating the synthesized voice quality. It is intended.

In order to achieve the above object, the present invention approximates a sound source amplitude control parameter obtained by analyzing a voice and a driving sound source cycle parameter between frames, and each parameter is an inflection point frame of this polygonal line. , The sound source amplitude control parameter value of each frame and the driving sound source cycle parameter value of each frame required at the time of synthesis are obtained by interpolation calculation from the parameter value of each inflection point frame adjacent to each other. This has the effect that information compression of the sound source amplitude control parameter and the driving sound source period parameter can be performed without deteriorating the noise.

Description of Embodiments Embodiments of the present invention will be described below with reference to the drawings. 4A to 4E are sound source amplitude control parameters (FIG. 4B), driving sound source cycle parameters (FIG. 4C) obtained by analyzing the same sound as shown in FIG. The approximation (Fig. 4D, E) and the original speech waveform (Fig. 4A) are shown. In the figure, indicates the inflection point. From this figure, the sound source amplitude control parameter and the driving sound source period parameter are approximated to a polygonal line, and only the frame at the inflection point has that value, so that not only the sound source amplitude control parameter but also the driving sound source period parameter information is obtained in the same manner. You can see that it can be compressed.

FIG. 5 shows a frame which must have a sound source amplitude control parameter value or a driving sound source period parameter value in the synthesis method according to the present invention. Further, in the figure, A indicates a frame in which a parameter value is calculated by interpolation calculation. Source amplitude control parameter is 30% of all frames
It can be seen that the driving sound source period parameter should have that value in about 30% of all frames.

EFFECTS OF THE INVENTION Since the present invention is the voice analysis method as described above, it is possible to perform the information compression of the sound source amplitude control parameter and the driving sound source cycle parameter without deteriorating the synthesized voice quality.

[Brief description of drawings]

FIG. 1 is a schematic diagram of a speech analysis and synthesis method, FIGS. 2A to 2C are waveform diagrams of an original speech waveform, a sound source amplitude control parameter, and a driving sound source period parameter in the conventional speech analysis and synthesis method, and FIG. 3 is a conventional method. 4A to 4E are explanatory diagrams of polygonal line approximation of the sound source amplitude control parameter and the driving sound source period parameter in the voice analysis and synthesis method in one embodiment of the present invention, and FIG. 5 is the same method. It is an explanatory view of a synthetic frame parameter value.

Claims (2)

[Claims] 1. A driving sound source amplitude control parameter obtained by analyzing a voice is approximated to a polygonal line between frames, and the driving sound source amplitude control parameter has only the value of an inflection point frame of this polygonal line, and each frame is driven at the time of synthesis. A voice analysis and synthesis method characterized in that a value obtained by interpolation calculation from values of adjacent inflection point frames is used as a sound source amplitude control parameter value. 2. A driving sound source cycle parameter obtained by analyzing speech is approximated to a polygonal line between frames, and the driving sound source cycle parameter has only the value of the inflection point frame of this polygonal line, and the driving sound source cycle of each frame at the time of synthesis. The voice analysis / synthesis method according to claim 1, wherein a value obtained by interpolation calculation from values of adjacent inflection point frames is used as the parameter value.