JPS5853359B2 - speech synthesizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a digital speech synthesizer, and particularly to a frequency parameter decoding method thereof.

A digital speech synthesizer analyzes and extracts sound source information and spectral envelope information that characterize the original speech, decodes parameters digitally encoded with low bits, and restores the original speech.

A conventional device of this type is shown in FIG.

In the figure, 1 is pitch data (reciprocal of frequency) conversion RO
M, 2 is a voiced pulse generator, 3 is an unvoiced pulse generator, 4 is a changeover switch, 5 is a multiplier, 6 is an amplitude conversion ROM
, 7 is a lattice type digital filter, 8 is a digital filter coefficient conversion ROM, 9 is a DA converter, and 10 is a speaker.

11 is a terminal to which a voiced sound pitch control input is applied; 12 is an input terminal for a signal that controls switching between voiced and unvoiced sounds; 13
14 is a sound source amplitude control input terminal, and 14 is a digital filter coefficient input terminal.

Next, the operation will be explained.

Among the voice parameters, encoded pitch control parameters of voiced sounds are applied to the pitch control input terminal 11 .

A voiced sound pulse having a period corresponding to the pitch data decoded by the pitch conversion ROM 1 is sent from the pulse generator 2 through the changeover switch 4, and encoded data input from the sound source amplitude control input terminal 13 is amplitude converted. R
The signal is decoded by the OM 6, and the amplitude is controlled by the multiplier 5 based on the amplitude data.

On the other hand, the unvoiced sound source from the unvoiced sound pulse generator 3 is sent to the multiplier 5 through the changeover switch 4.

The sound source signal whose amplitude has been controlled by the multiplier 5 is supplied to a lattice type digital filter 7.

The coefficients of the digital filter are encoded and inputted from the digital filter coefficient input terminal 14, and are input to the conversion ROM 8.
controlled by the data decoded by

The output of the digital filter 7 is passed through a DA converter 9, and a speaker 10 generates audio.

By the way, the pitch parameter of the sound source information is generally given as an integral multiple of the audio sampling period, and the sampling frequency is set to 10
If KHz is used, a normal woman would have 20 to 70 pitches (14
0-500H2), 40-140 pitches (70
~250H2), and approximately 7 bits are required to quantize pitch data including men and women.

Even if non-linear quantization is performed that takes a large quantization step at low frequencies, about 6 bits are required to prevent sound quality deterioration, and the conversion ROM capacity at that time is 2' = 6.
It required 4 words.

■A system in which words are 10 bits has the drawback of requiring a large ROM capacity of 640 bits, and the amount of information for pitch parameters is 640 bits with a frame period of 10 MS.
Even in the case of 00 bits/sec and 20M5, there was a drawback that 300 bits/sec was required.

This invention was made in order to eliminate the above-mentioned drawbacks of the conventional technology, and it is possible to reduce the pitch period conversion ROM capacity by providing a means for multiplying the data in the pitch parameter conversion ROM by a constant based on information specifying the frequency band. Moreover, it is an object of the present invention to provide a speech synthesis device that can reduce the amount of information.

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

In Figure 2, a circuit that doubles 15 pitch data, 1
6 is an input terminal for a signal specifying a frequency band.

The parts numbered 1 to 14 are the same as in FIG. 1, and their explanation will be omitted.

Among the voice parameters, pitch control parameters for voiced sounds are input to the pitch control input terminal 11 .

The pitch data decoded by the pitch conversion ROMI is sent to a circuit 15 that can double it, and the data multiplied by 2 or 1 depending on the information at the input terminal 16 specifying frequency band information is sent to the voiced pulse generator. 2 to generate voiced sound pulses corresponding to the pitch data.

For example, assume that the pitch conversion ROM contains pitch data corresponding to women.

A pitch range of 20 to 70 is sufficient for women, so 6
It is sufficient to quantize in bits, and even if the quantization step is increased at low frequencies, there is little deterioration in sound quality, so about 5 bits is sufficient.

It is sufficient for the male pitch range to be about twice that of the female, and this is obtained by the circuit 15 which doubles the pitch data using the input terminal 16 that specifies the frequency band.

A doubling circuit can be easily realized by a shift operation or the like.

In this example, the pitch conversion ROM capacity is 25 = 32 words, and the amount of information of the pitch parameter is a frame period of 10M5.
This can be reduced to 500 bits/sec for 20M5 and 250 bits/sec for 20M5, compared to the conventional technology.

In the above embodiment, a doubling circuit based on a shift operation is used as means for multiplying pitch data by a constant, but a 1/2 circuit or a multiplier for multiplying by an arbitrary coefficient may also be used.

As described above, according to the present invention, there is provided a means for multiplying the data in the pitch parameter conversion ROM by a constant based on the information specifying the frequency band in the speech synthesizer.
This has the effect of reducing the M capacity and the amount of information.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a conventional speech synthesis device, and FIG. 2 is a block diagram showing a speech synthesis device according to an embodiment of the present invention. 1 is a pitch parameter conversion ROM, 15 is a constant multiplication circuit, and 16 is an input terminal for specifying a frequency band.

Claims (1)

[Claims] 1. Encode and transmit the voice feature parameters and convert R
1. A digital speech synthesis device using parameters decoded by an OM, characterized in that the device includes means for multiplying data in a pitch parameter conversion ROM by a constant based on information specifying a frequency band.