JPH03174597A - Voice synthesizer - Google Patents

Abstract

PURPOSE:To generate a more natural synthesized voice by multiplying a synthesized voice signal by the output of a fluctuation time-series memory through an integrator and adding amplitude fluctuations to the synthesized voice. CONSTITUTION:The fluctuation time series memory 2a is stored with amplitude fluctuations obtained from a natural voice by an amplitude fluctuation extracting means 1 in advance. For example, a vowel which is generated steadily is used as the natural voice and the amplitude fluctuation extracting means 1 finds a maximum amplitude value in each pitch period of the natural voice and divide maximum values by a mean value for normalization to obtain amplitude fluctuations. The voice signal outputted by a synthesizing filter 4 is superposed upon the amplitude fluctuations from a fluctuation generating circuit 2 to obtain a voice signal given the amplitude fluctuations. Consequently, the amplitude fluctuations in the natural voice can be reflected faithfully and the more natural synthesized voice can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a speech synthesis device, and more particularly, to the configuration of a device for imparting naturalness to one synthesized speech in a regular speech synthesis device.

2. Description of the Related Art Conventional regular speech synthesis devices control the amplitude of synthesized speech according to predetermined rules. for example.

In the same way as obtaining pitch patterns based on prosodic information,
There is a method for obtaining an amplitude pattern.

By the way, since the human voice is uttered by a living body, it contains various fluctuations, one of which is amplitude fluctuation. Therefore, conventionally, this amplitude fluctuation is approximately synthesized and superimposed on the amplitude pattern. Various methods have been proposed for approximate synthesis of this amplitude fluctuation. For example, (a) Tokuko Sho 62-4
In Publication No. 9639, the output of a normal random number generation means is used as a fluctuation sequence (this is a method of adding fluctuation data to an audio signal). (b) Japanese Patent Publication No. 63-229
No. 499 uses an integral filter for the output of the random number generating means to obtain a fluctuation sequence (this is a method of integrating the fluctuation sequence plus a constant and the audio signal). (c) In JP-A-58-186800, a 1/f fluctuation is given to the amplitude pattern (no information is given on how to give the fluctuation)a Also, (
d) In Japanese Patent Laid-Open No. 55-133099, fluctuations obtained from an amplifier are added to the reference voltage of a D/A converter. In the prior art (a), the output of the random sequence oscillator is converted into a normal random number sequence and used as the amplitude fluctuation. Also, in (b), by passing the random sequence through an integrator.

Amplitude fluctuations are obtained. In addition, in (c), the random sequence is 1/
Approximation of f fluctuation is performed. Furthermore, in (d), the output fluctuation of an amplifier with a very large amplification factor is used. However, since the above-mentioned conventional method was only an approximation of amplitude fluctuation, unnaturalness remained.

1-One Attack The present invention was made in view of the above-mentioned circumstances,
In particular, this was done with the aim of reflecting the amplitude fluctuations of the living body to obtain more natural synthesized sounds.

In order to achieve the above object, the present invention provides a regular speech synthesizer that determines the amplitude of synthesized speech based on certain rules or stored data, which includes a fluctuation time series memory;
The apparatus is characterized in that it has an integrator and adds amplitude fluctuation to one synthesized speech by multiplying the synthesized speech signal and the output of the fluctuation time series memory by the integrator. Hereinafter, the present invention will be explained based on examples.

FIG. 3 is a circuit configuration diagram for explaining an example of an amplitude fluctuation generation circuit used in the present invention. In the figure, 1 is an amplitude fluctuation extraction circuit, 2 is an amplitude fluctuation generation circuit, and the amplitude fluctuation generation circuit 2 is It has a fluctuation time series memory 2a and an address counter 2b, and the fluctuation time series memory 2a stores fluctuations digitally. This fluctuation time series memory 2a is addressed, for example, by a D-bit address counter 2b. The address counter 2b is incremented every pitch period by the pitch period signal. In another embodiment, the address counter may control a clock oscillator with a pitch period signal.

The amplitude fluctuation obtained from the real voice by the amplitude fluctuation extraction means 1 is stored in advance in the fluctuation time series memory 2a. For example, using a regularly uttered vowel voice as the real voice, the amplitude fluctuation extraction means 1 extracts one of the real voices.
The maximum value of the amplitude is determined for each pitch period, and the value is normalized by dividing it by the average value, and the result is defined as the amplitude fluctuation.

FIG. 1 and FIG. 2 are block diagrams for explaining embodiments of a speech synthesis device according to the present invention, respectively.

In the figure, 2 is an amplitude fluctuation generation circuit, 3 is a sound source generation section, 4 is a synthesis filter, and 5 is an integrator.In the embodiment shown in FIG. 5, the signal is superimposed with the amplitude fluctuation from the fluctuation generation circuit 2, resulting in an audio signal given amplitude fluctuation. As another example, as shown in FIG. 2, the same effect can be obtained even if amplitude fluctuation is applied to the sound source signal. Further, in a waveform editing type speech synthesis device, a similar effect can be obtained by superimposing one synthesized speech waveform and the above amplitude fluctuation using an integrator.

It is also possible to store a plurality of different fluctuations in the fluctuation time-series memory and switch them for each phoneme to be synthesized, for example, to provide more natural fluctuations.

As is clear from the above description, according to the present invention, the fluctuation generation circuit of the speech synthesizer of the present invention can faithfully reflect the amplitude fluctuations included in the real voice, making it more natural. It is possible to obtain high-quality synthesized speech.

[Brief explanation of the drawing]

FIGS. 1 and 2 are block diagrams for explaining the present invention in detail, and FIG. 3 is a block diagram for explaining an example of an amplitude fluctuation generation circuit used to implement the present invention. DESCRIPTION OF SYMBOLS 1... Amplitude fluctuation extraction circuit, 2... Amplitude fluctuation generation circuit, 2a... Fluctuation time series memory, 2b... Address counter, 3... Sound source generation unit, 4... Synthesis filter, 5 ...Integrator.

Claims (1)

[Claims] 1. A regular speech synthesizer that determines the amplitude of synthesized speech based on certain rules or stored data, which has a fluctuation time series memory and an integrator, and which outputs the synthesized speech signal and the fluctuation time series memory. , a speech synthesis device characterized in that the synthesized speech is multiplied by the integrator to impart amplitude fluctuation to the synthesized speech.