JPS62102300A - Voice synthesizer - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a speech synthesizer, and in particular extracts basic speech parameters including spectral information and sound source information from speech, quantizes them, and then transmits them. The present invention relates to a speech synthesizer in a speech analysis and synthesis device that enables narrowband compressed transmission of speech by resynthesizing the original speech.

[Conventional technology]

Traditionally, this type of speech synthesizer has been designed with 20 frames called frames.
Spectral parameters such as partial autocorrelation functions are received at regular intervals of about milliseconds, and the spectrum is transmitted at intervals shorter than the frame time called subframes to eliminate discontinuity with the previous and subsequent frames. The parameters were interpolated between the spectral parameters of the previous and subsequent frames and used as coefficients of the synthesis filter.

[Problem that the invention seeks to solve]

The interpolation function in the conventional speech synthesizer described above is effective when there is little difference in the predicted gain of the spectral parameter between the previous and subsequent frames, but when there is a significant change in the predicted gain of the spectral parameter between the previous and subsequent frames. This method has the disadvantage that the prediction gain of the synthesis filter does not change linearly in the subframe during interpolation, resulting in a decrease in synthesized sound quality or the generation of noise due to oscillation of the synthesis filter.

[Means for solving problems]

The speech synthesizer of the present invention is equipped with a means for stopping sysvector interpolation based on a change in the predicted gain of the received spectral parameter, and calculates the predicted gain from the received spectral parameter. means for calculating, means for determining the absolute value of the difference in prediction gain between frames, means for comparing the absolute value of the difference in prediction gain with a predetermined threshold, and a comparison result obtained by this means. It has means for stopping spectral parameter interpolation.

[Effect]

A change in the predicted gain of the received spectral parameter is detected, and when the change value becomes larger than a predetermined threshold, interpolation of the spectral parameter is stopped.

〔Example〕

Embodiments of the present invention will be described in detail below based on the drawings.

The figure is a block diagram showing an embodiment of a speech synthesizer according to the present invention.

, In the figure, 1 is a spectral parameter input terminal into which a spectral parameter is input, 2 is a sound source signal input terminal into which a sound source signal is input, 3 is a prediction gain calculation circuit that calculates a prediction gain from the received spectral parameter, and 4 is a prediction gain calculation circuit. A register for storing the prediction gain of the previous frame; 5 a differentiator for calculating the difference between the prediction gain of the previous frame stored in this register 4 and the prediction gain from the prediction gain calculation circuit 3; 6;
7 is an absolute value circuit which takes the output of this subtractor 5 as input and calculates the absolute value of the difference in predicted gain between frames, and 7 compares the absolute value of the difference in predicted gain from the absolute value circuit 6 with a predetermined threshold value. 8 is a register in which the spectral parameters of the previous frame are stored for interpolation. 9 is an interpolator that receives the output of this register 8 and the spectral parameters from the spectral parameter input terminal 1. This interpolator 9 is constant. The spectral parameters received at time intervals are interpolated to generate interpolated spectral parameters. 10 is the output of this interpolator 9 and the sound source signal input terminal 2
11 is an audio output terminal from which a synthesized sound is obtained.

Next, the operation of the embodiment shown in this figure will be explained.

First, spectral parameters input manually through the spectral parameter input terminal 1 are input to the predicted gain calculation circuit 3, and the predicted gain is calculated. The prediction gain obtained by the prediction gain calculation circuit 3 and the prediction gain of the previous frame stored in the register 4 are subjected to a difference between them by a subtractor 5, and then input to an absolute value circuit 6.

Next, the change in the spectral parameter that is the output of the absolute value circuit 6 is compared with a predetermined threshold value in a comparator 7, and if it exceeds the predetermined threshold value, it is sent to the interpolator 9 as an interpolation stop signal a. is input. The spectral parameters selectively interpolated by the interpolator 9 are input as filter coefficients, and the sound source signal input from the sound source signal input terminal 2 is input as a filter drive signal to the synthesis filter 10, and the synthesized sound is output from the audio output terminal 11.

〔Effect of the invention〕

As explained above, according to the present invention, a change in the predicted gain of a received spectrum parameter is detected without using complicated means, and when the change value becomes larger than a predetermined threshold, the spectrum parameter A simple configuration in which interpolation is stopped can prevent amplitude abnormalities in the synthesized sound and provide higher quality speech, so the practical effects are extremely significant.

[Brief explanation of drawings]

The figure is a block diagram showing one embodiment of the present invention. 3...Predicted gain calculation circuit, 5...Differentiator, 6...
... Absolute value circuit, 7 ... Comparator, 9 ... Interpolator, 10 ... Synthesis filter.

Claims (1)

[Claims] A speech synthesizer includes means for interpolating spectral parameters received at regular time intervals to generate interpolated spectral parameters, and means for synthesizing speech using the received driving sound source signal and the interpolated spectral parameters. A speech synthesizer, comprising means for stopping spectral parameter interpolation due to a change in prediction gain.