JPH0363760B2 - - Google Patents

Description

[Detailed description of the invention] The present invention relates to a speech synthesis device.

Conventionally, a speech synthesis method is known in which speech is synthesized using a formaton (polar) parameter as a spectral envelope parameter of speech and pitch data, voiced and unvoiced data, and amplitude data as sound source parameters. Also, using the above method,
A one-chip speech synthesis LSI is known that is implemented as an LSI together with a synthesis data memory that stores synthesis data such as spectral envelope parameter values and sound source parameter values.

However, the conventional method is a relatively rough model (so-called all-pole model) that approximates the spectral envelope characteristics of speech using the characteristics of multiple poles, and the sound source waveform is also a plosive sound source. However, due to factors such as the lack of a voice, the quality of the synthesized speech is not sufficient.

An object of the present invention is to provide a speech synthesis device that is relatively small in scale, can be implemented in LSI, and can generate synthesized speech of relatively high quality.

The present invention provides a speech synthesis device of the type that synthesizes speech waveforms using synthesis data such as pole-zero parameters, pitch cycle data, and voiced and unvoiced data, in which the connection configuration of each of the circuits including a pole circuit and a zero circuit is variable. a synthesis filter, a sound source waveform generation circuit that generates a plurality of selectable types of sound source waveforms, means for selecting a connection configuration of the synthesis filter and the sound source waveform generation circuit according to synthesis data, and synthesis data that stores the synthesis data. It is comprised of a memory and means for selecting or editing the synthesis data in accordance with a voice output command given as an input to generate a desired voice synthesis waveform.

The feature of the present invention is that by making the configuration of a plurality of polar circuits and zero circuits variable according to phonetic data, etc., it is possible to perform real-time processing even on a relatively small scale and with a relatively slow calculation speed. By making it possible to select multiple types of sound source waveforms, such as a sound source generation circuit, a noise generation circuit, a plosive sound source generation circuit that generates a plosive sound source waveform, and means for supplying external sound source waveforms such as predicted residual waveform data, relatively high quality can be achieved. It is possible to obtain a speech synthesizer capable of generating synthesized speech with good quality. Furthermore, the present invention can be used as a so-called analysis-synthesis type speech synthesizer that stores and uses synthesized data obtained by analyzing natural speech waveforms, and also can generate synthesized data from character sequences given as input. It is also possible to use it as a so-called rule synthesis type speech synthesizer that automatically generates speech.

Next, the present invention will be explained in detail using the drawings.

First, FIG. 1 is a block diagram showing an embodiment of the speech synthesis apparatus of the present invention. Voice output command data such as character series or text number data is input to the control circuit 6 via the voice output data input terminal 1. The control circuit 6 generates address data for the composite data according to the audio output command data, and outputs it to the composite data storage circuit 2 via the address data transmission line 3. The composite data storage circuit 2 outputs composite data such as spectral envelope parameter values, sound source data, etc. to the control circuit 6 via the composite data transmission path 4 in accordance with the address data.

Next, the control circuit 6 edits the synthesis data and outputs it to the interpolation circuit 9 via the synthesis data transmission line 7, and also generates configuration data for the speech synthesis circuit according to the synthesis data. The data is output to the configuration control circuit 12 via the configuration data transmission line 8. The interpolation circuit 9 interpolates the composite data and outputs the interpolated value for the spectral envelope parameter value as the control parameter value of the polar circuit and the zero circuit to the composite filter unit 21 via the composite filter control data transmission path 11. The interpolated value for the sound source data is output to the sound source waveform generation section 15. On the other hand, the configuration control circuit 12 synthesizes configuration data of polar circuits and zero circuits via a synthesis filter configuration data transmission line 14 according to the configuration data. Filter section 21
At the same time, sound source waveform selection data is generated according to the configuration data and output to the sound source waveform generation section 15 via the sound source waveform selection data transmission line 13. The sound source waveform generation unit 15 includes an external sound source generation circuit 16 that generates a sound source waveform in accordance with the sound source waveform data inputted from the synthetic data storage circuit 2 via the sound source waveform data transmission line 5 in accordance with the selection data of the sound source waveform; A triangular wave sound source generation circuit 17 that generates a triangular wave as a sound source waveform, a noise waveform generation circuit 18 that generates a noise waveform as a sound source waveform,
One of the plosive sound source generation circuits 19 that generate a plosive sound source is selected, and a sound source waveform is generated according to the interpolated value for the sound source data and output to the synthesis filter section 21. The synthesis filter section 21 converts polar circuits 22, 24,
25, 27 zero circuits 23, 26 and adder circuit 28
A synthesized waveform is generated according to the interpolated value of the spectral envelope parameter value and the sound source waveform, and is outputted via the synthesized waveform output terminal 29.

Next, first and second embodiments of the sound source generation section and the synthesis filter section will be described.

FIG. 2a is a block diagram showing an example of the configuration when synthesizing nasal sounds as a first embodiment.

In the sound source generation section, a triangular wave sound source generation circuit 20
1 is selected, and the triangular wave sound source generation circuit 201 generates a triangular waveform every pitch period according to the sound source data provided via the sound source data transmission line 202 and outputs it to the synthesis filter section 203. Synthesis filter section 2
03, the pole circuits 205, 207, 208 and the zero circuit 206 are continuously connected, and each pole circuit and zero circuit is controlled according to the synthesis filter control data given via the synthesis filter control data transmission line 204, and the triangular A composite waveform is generated using the waveforms and outputted via the composite waveform output terminal 209.

FIG. 2b is a block diagram showing an example of a configuration when synthesizing a voiced masatsu sound (for example, a consonant part of a za sound) as a second embodiment.

In the sound source generation section 210, the noise source generation circuit 2
12 and a triangular wave sound source generation circuit 213 are selected and generate noise and a triangular wave according to the sound source data provided via the sound source data transmission line 211, respectively, and output them to the synthesis filter section 214. In the synthesis filter section 214, the cascade-connected polar circuits 2
17, 218, 219 and the polar circuit 216 are connected in parallel. The polar circuit 216 is a triangular wave, the polar circuit 217 is
218 and 219 are each excited by noise,
The outputs are added together by an adder circuit 220 and output as a composite waveform via a composite waveform output terminal 221.

In the above embodiments, both a and b in the figure are constructed using four pole circuits or zero circuits, and the amount of calculation required in the synthesis filter section is almost the same. As is clear from this, according to the present invention, it is possible to realize various circuit configurations without increasing the amount of calculations, and it is possible to generate synthesized sounds of relatively high quality.

According to the present invention, by making the configuration of the synthesis filter consisting of polar circuits and zero circuits variable, it is possible to realize various circuit configurations with a relatively small amount of calculation, and also to realize multiple circuit configurations including plosive sound sources and external sound sources. It is clear that relatively high quality synthesized speech can be obtained by selectively using the sound source circuits.

Furthermore, it is also possible to implement it on one chip using the large-scale integrated circuit (LSI) technology of the present invention.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention;
Figures 2a and b are the sound source waveform generator 1 in Figure 1.
5 is a block diagram showing concrete first and second embodiments of the synthesis filter section 21. FIG. 1 is an audio output data input terminal, 2 is a synthetic data storage circuit, 3 is an address data transmission line, 4 is a synthetic data transmission line, 5 is a sound source waveform data transmission line, 6 is a control circuit, 7 is a synthetic data transmission line, 8 is a configuration data transmission line, 9 is an interpolation circuit, 1
0 is a sound source data transmission line, 11 is a synthesis filter control data transmission line, 12 is a configuration control circuit, 13 is a sound source waveform selection data transmission line, 14
15 is a synthesis filter configuration data transmission line, 15 is a sound source waveform generation section, 16 is an external sound source generation circuit, 17 is a triangular wave sound source generation circuit, 18 is a noise source generation circuit, 19 is a plosive sound source generation circuit, and 20 is a sound source Waveform transmission line, 21 is a synthesis filter section, 22, 2
4, 25, 27 are polar circuits, 23, 26 are zero circuits,
28 is an adder circuit, and 29 is a composite waveform output terminal. 201 and 213 are triangular wave sound source generation circuits, 20
2, 211 is a sound source data transmission line, 203, 214
is a synthesis filter section, 204, 215 is a synthesis filter control data transmission line, 205, 207, 208,
216, 217, 218, 219 are polar circuits, 20
6 is a zero circuit, 209 and 221 are composite waveform output terminals, 212 is a noise source generating circuit, 220 is an adder circuit,
It is.

Claims (1)

[Claims] 1. In a speech synthesis device of the type that synthesizes a speech waveform using synthesis data such as pole-zero parameters, pitch cycle data, voiced and unvoiced data, etc., a synthesis filter including a pole circuit and a zero circuit and having a variable connection configuration of each of the circuits is provided. , a sound source waveform generation circuit that generates a plurality of selectable types of sound source waveforms, means for selecting a connection configuration of the synthesis filter and the sound source waveform generation circuit according to the synthesis data, and a synthesis data memory that stores the synthesis data. and means for selecting or editing the synthesis data in accordance with a voice output command given as an input to generate a desired voice synthesis waveform.