JPH07311595A - Voice synthesizer - Google Patents

Abstract

PURPOSE:To generate a synthesized voice having a natural and enjoyable tone quality by inserting a specific filter between a driving sound source and a vocal tract function filter. CONSTITUTION:The voice waveforms inputted to an analysis section 1 are accumulated in a buffer 3 and are made into blocks for every prescribed length. The voices, which are made into blocks, are transmitted to each of a vocal tract analysis section 4, a voiced.unvoiced analysis section 5 and a pitch extracting section 6, respectively. The pitch period obtained by the section 6 is inputted to a pulse sound source generating section 7 and impulse train having specific intervals is generated. Then, the pulse train is inputted to a 1/f<n> (where n is a positive real number) filter 8, filtered and outputted. Based on the 'voiced' and 'unvoiced' discrimination result of the section 5, a changeover of a switch 9 is conducted and the output of the 1/f<n> filter 8 is inputted to a synthesis filter 11 as a driving sound source in the case of 'voiced', and the noise sound source output of a noise sound source 10 is inputted in the case of 'unvoice'.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice synthesizing device for outputting a synthetic voice synthesized according to an input voice waveform, and more particularly to a voice synthesizing device capable of generating a synthetic voice having a natural and easy-to-listen sound quality.

[0002]

2. Description of the Related Art Generally, in a conventional speech synthesizer of a type in which a vocal tract function filter is driven by a sound source, noise (unvoiced sound) and a pulse train as a sound source or a derivative of a vocal tract function is applied to a speech signal. Representative residuals are used which are classified based on their characteristics. However, in a voice synthesized in this manner, the low frequency components are deteriorated as compared with the real voice, so that the synthesized voice is mechanical, lacks in naturalness, and is aurally unnatural. There was a problem.

[0003]

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a voice synthesizing device capable of generating a synthetic voice having a sound quality that is natural and easy to hear.

[0004]

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention provides a voice synthesizer for driving a vocal tract function filter by a driving sound source to synthesize a voice, wherein 1 / is provided between the driving sound source and the vocal tract function filter. It is characterized in that a filter of f ⁿ (n is a positive real number) is inserted. Another feature of the present invention has an analysis unit that analyzes an input speech waveform to obtain a speech parameter, and a synthesis unit that drives a vocal tract filter with a driving sound source to obtain an output synthesized speech waveform from the speech parameter. In the voice synthesizer, a 1 / √f filter is provided between the driving sound source and the vocal tract function filter in the synthesizer.

[0005]

The present invention will be described below based on the illustrated embodiments. FIG. 1 is a block diagram showing an embodiment of a speech synthesizer according to the present invention. As shown in FIG. 1, this speech synthesizer includes a buffer 3 into which an input speech waveform is input, a vocal tract analysis unit 4, a voiced / unvoiced analysis unit 5, and a pitch extraction unit 6 which are connected to the buffer 3, respectively. A pulse sound source generation unit 7 connected to the pitch extraction unit 6; a 1 / √f filter 8 connected to the pulse sound source generation unit 7;
/ √f filter 8 and a switch 9 connected to the voiced / unvoiced analysis unit 5, a noise sound source 10 connected to the switch 9, a synthesis filter connected to the switch 9 and the vocal tract analysis unit 4 (voice Road function filter) 11.

The buffer 3, the vocal tract analyzer 4,
The voiced / unvoiced analysis unit 5 and the pitch extraction unit 6 constitute the analysis unit 1, and the pulse sound source generation unit 7, 1 / √f
The filter 8, the switch 9, the noise sound source 10, and the synthesis filter 11 constitute the synthesis unit 2.

Next, the operation of the speech synthesizer having the above configuration will be described. First, when viewed broadly, the analysis unit 1 analyzes the input speech waveform to obtain a speech parameter, and the synthesis unit 2 obtains an output synthesized speech waveform from the speech parameter obtained by the analysis unit 1.

Next, the individual detailed operations will be described. The input voice waveform input to the analysis unit 1 is stored in the buffer 3 and is divided into blocks of a fixed length. The voices blocked by the buffer 3 are respectively passed to the vocal tract analysis unit 4, the voiced / unvoiced analysis unit 5, and the pitch extraction unit 6, and the vocal tract analysis unit 4 performs frequency analysis to obtain vocal tract information. Obtained, and in the voiced / unvoiced analysis unit 5,
Whether the input voice is voiced / unvoiced is determined, and the pitch extraction unit 6
In, when the input voice is voiced, the fundamental frequency (pitch frequency) is analyzed and the pitch period is output. The outputs of the vocal tract analysis unit 4, the voiced / unvoiced analysis unit 5, and the pitch extraction unit 6 are supplied to the synthesis unit 2.

The pitch cycle obtained by the pitch extraction section 6 is inputted to the pulse sound source generation section 7 and an impulse train having an interval corresponding to the pitch cycle as shown in FIG. 2 is generated. Next, this pulse train is the 1 / √f filter 8 described above.
Is input to, and filtered as shown in the following equation (1) and output. Eφ (o) = √π Eφ (i) = {(2i−1) / 2i} Eφ (i−1) (1 ≦ i ≦ N) (1) The output waveform from the 1 / √f filter 8 is The result is shown in Fig. 3. As shown in FIG. 3, the pulse train generated here has an attenuation characteristic of 1 / √f in the section of Eφ (i). That is, it can be said that the filter 8 plays a role of a 1 / f ⁿ (n is a positive real number) filter.

Then, the voiced / unvoiced analysis part 5
The switch 9 is switched according to the result of the unvoiced judgment. In the case of voice, the output of the 1 / √f filter 8 is
In the case of no voice, the noise sound source output of the noise sound source 10 is input to the synthesis filter (vocal tract function filter) 11 as a driving sound source. The synthesis filter 11 has the vocal tract information obtained by the vocal tract analysis unit 4 as a coefficient, and obtains and outputs a synthetic speech waveform by the driving sound source.

With this configuration, in the voiced section,
Since the low frequency band is emphasized on the frequency component of the synthetic voice, it is possible to synthesize a synthetic voice having a natural and easy-to-listen sound quality. That is, for example, as a result of an experiment using the voice phrase "Aishinari", the conventional device obtains the synthesis result shown in FIG. 4, whereas the device of the present invention produces the result shown in FIG.
A clear (shaded) synthetic speech was obtained as shown in.

[0012]

As described above, according to the present invention, in a voice synthesizer for driving a vocal tract function filter (synthesis filter) with a driving sound source, a 1 unit is provided between the driving sound source and the vocal tract function filter.
Insert a filter of / f ⁿ (n is a positive real number), and sound source → 1
/ F ⁿ filter → Since the configuration of the vocal tract function filter,
Since the low frequency band is emphasized on the frequency component of the synthesized voice in the voiced section, it is possible to synthesize a synthesized voice having a natural and easy-to-listen sound quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a speech synthesizer according to the present invention.

2 is a waveform diagram showing an impulse train generated in the pulse sound source generation unit shown in FIG.

FIG. 3 is an output waveform diagram from the 1 / √f filter shown in FIG.

FIG. 4 is a diagram showing a synthesized voice result in a conventional device.

FIG. 5 is a diagram showing a synthesized voice result in the device of the present invention.

[Explanation of symbols]

1 ... Analysis unit, 2 ... Synthesis unit, 3 ... Buffer, 4 ... Vocal tract analysis unit, 5 ... Voiced / unvoiced analysis unit,
6 ... Pitch extraction unit, 7 ... Pulse sound source generation unit,
8 ... 1 / √f filter, 9 ... switch,
10 ... Noise source, 11 ... Synthesis filter (vocal tract function filter),

Claims (2)

[Claims] 1. A voice synthesizer for synthesizing a voice by driving a vocal tract function filter with a driving sound source, wherein 1 / f ⁿ (n is a positive real number) between the driving sound source and the vocal tract function filter. A voice synthesizer characterized by inserting the filter of. 2. A speech synthesizer having an analyzing section for analyzing an input speech waveform to obtain a speech parameter, and a synthesizing section for driving a vocal tract function filter with a driving sound source to obtain an output synthesized speech waveform from the speech parameter. And 1 between the driving sound source and the vocal tract function filter in the synthesis unit.
A voice synthesizing device provided with a / √f filter.