JP2956936B2 - Speech rate control circuit of speech synthesizer - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an utterance speed control circuit of a speech synthesizer,
In particular, the present invention relates to an utterance speed control circuit that outputs a synthesized sound that is natural and easy for a listener to hear even when the utterance speed is changed in a speech synthesizer.

[Problems to be Solved by Conventional Techniques and Inventions] Conventional speech synthesizers are, for example, as described in Japanese Patent Application Laid-Open No. 50-153807, such as / p /, / t /, / k /. The expansion and contraction rate is reduced for short bursting consonants, and the expansion and contraction rate is increased for stable vowels to control the utterance speed of synthesized sounds in units of phonemes. However, simply changing the expansion / contraction ratio for each phoneme does not sufficiently consider the preservation of fine phonological properties, and has a disadvantage that unnaturalness remains when a synthesized sound is heard.

SUMMARY OF THE INVENTION Therefore, a main object of the present invention is to provide a speech rate control circuit that can generate a synthesized speech that is natural and easy for a listener to hear even when the speech rate is changed in a speech synthesizer. .

[Means for Solving the Problems] The utterance according to the first aspect is obtained by converting input means for inputting voice, A / D converting means for converting input voice into a digital signal, and converting the voice. Analysis means for analyzing a digital speech signal to extract feature parameters; synthetic speech conversion means for converting the extracted feature parameters into synthesized speech; and D / A for converting the converted synthesized speech into an analog signal and outputting the analog signal A spectrum analysis means for converting the converted digital voice signal into a spectrum parameter carrying voice information, and a spectrum change for obtaining a temporal rate of change using the converted spectrum parameter. Utterance speed control for controlling a utterance speed by determining a synthetic frame period of a feature parameter based on the analysis means and the obtained temporal change rate Means for converting the synthesized speech signal into and out of time based on the determined synthesized frame period.

According to a second aspect of the present invention, there is provided input means for inputting a character string or a symbol string, reading string conversion means for converting an input character string or a symbol string into a reading string, and converting the converted reading string. In a speech synthesizer comprising a synthesized speech conversion means for converting into a synthesized speech, and a D / A conversion means for converting the converted synthesized speech into an analog signal, the converted read sequence is converted into a spectrum parameter carrying audio information. A spectrum analysis means for converting, a spectrum change analysis means for obtaining a temporal change rate using the converted spectral parameters, and a utterance by determining a composite frame period of the characteristic parameter based on the obtained temporal change rate Speech rate control means for controlling the rate, wherein the synthesized speech conversion means performs time expansion and contraction of the synthesized speech signal based on the determined synthesized frame period. One in which the.

[Operation] The utterance speed control circuit of the speech synthesizer according to the present invention comprises:
Convert the audio signal or the converted read sequence into parameters related to the spectrum, determine the temporal change rate using the spectral parameters, determine the composite frame period of the feature parameter based on the determined temporal change rate The utterance speed is controlled, and the synthesized speech signal is time-expanded and output based on the determined synthesized frame period.

FIG. 1 is a schematic block diagram of an embodiment of the present invention.

In FIG. 1, a voice input unit 10 receives a uttered voice, and the uttered voice is input to an A / D conversion conversion unit 1 and converted into a digital signal by a sampling signal at a predetermined interval. Is converted to The voice converted into the digital signal is supplied to the spectrum analysis unit 3 and is converted into a spectrum parameter carrying phonological information.
The spectral parameters include, for example, PARCOR coefficients described in “Basics of speech information processing” by Saito and Nakata (published by Ohmsha). This PARC
The OR coefficient is an example of a spectrum parameter, and the present invention is not particularly limited to this.

The voice digital signal output from the A / D conversion unit 1 is also input to the voice analysis unit 2, and sound source parameters such as pitch period and power information as sound source information are extracted from the voice digital signal. The sound source parameters extracted by the sound source analysis unit 2 and the spectrum parameters analyzed by the spectrum analysis unit 3 are provided to a synthesis unit 5. The synthesis unit 5 generates a synthesized speech using the spectrum parameters and the sound source parameters.

That is, the synthesizing unit 5 repeats the spectrum parameter in the unit frame at the above-described pitch period interval, and converts the speech parameter sequence. As the synthesizing unit 5, for example, a two-multiplier lattice type speech synthesizing filter described in “Linear Prediction of Speech” (published by Corona) by JDMarkel and AHGray Jr is used. And synthesize a speech waveform that is good for the sound source parameters. Note that the squaring multiplier-type speech synthesis filter is an example, and other speech synthesis means may be used. The synthesized voice waveform is output as voice via the D / A converter 7.

Next, the control of the utterance speed of the synthesized speech will be described. The control of the utterance speed of the synthesized voice is realized by changing the synthesis frame period of the sound source pulse and the spectrum parameter when the synthesizer 5 synthesizes the voice waveform from the sound source parameter and the spectrum parameter. The utterance control is performed by the spectrum change analyzer 4 and the utterance speed controller 6. First, in the spectrum change analysis unit 4, the spectrum analysis unit 3
The temporal change rate Δ is obtained using the spectrum parameters calculated in the above. For example, Sagayama, Itakura,
The dynamic measure using the LPC cepstrum regression coefficient, as described in the “Personality information included in the dynamic measure of speech”, Proceedings of the Acoustical Society of Japan (June 1979) Can be used. Of course, the scale of the spectrum change rate is not limited to this.

FIG. 2 (a) shows an example of the time change of the voice spectrum change rate Δ, and FIG. 2 (b) shows the average value of the change rate Δ for each frame. FIG. 2 (c) shows an example of sound source pulses and spectral parameters when uttering at the same speed as the original voice. In FIG. 2 (c), the sound sources of the i-th frame and the j-th frame are displayed. Shown. Also, the spectral parameters in each frame are represented by vectors ｛P ⁱ ₁ and P ⁱ , respectively.
₂ ... P ⁱ _m}, is represented by _{^{{P j 1, P j 2}} ... P j m}.

Next, a case in which the utterance speed of the synthesized speech is reduced, that is, the speech is extended in the time axis direction will be described. In one embodiment of the present invention, as shown in FIG. 2 (b), the composite frame period of the characteristic parameter is determined based on the spectrum change rate Δ obtained for each frame. In the example shown in FIG. 2B, the spectrum change rate Δi of the i-th frame, and in the example shown in FIG. ,
A new synthesized frame period LNi is determined by the following equation (1).

LNi = Li × Δmax / Δi (1) where It is. In this way, an inverse phase relationship is provided between the magnitude of the value of the spectrum change rate Δ and the expansion / contraction rate to newly generate the sound source pulse and the spectrum parameters {P ⁱ ₁ , P ^{i as} shown in FIG. ₂ … ^Pim _m is placed.

The above procedure can also be applied to the case where the synthesized speech utterance speed is increased. Equation (1) is set as LNi = Li × Δi / Δmax (2) and a new synthesized frame period LNi is determined. Here, the relationship between the spectrum change rate and the expansion / contraction rate need not be an inverse correlation, and an arbitrary function can be used.

FIG. 3 is a schematic block diagram showing another embodiment of the present invention.

In FIG. 3, to an input terminal 20, a sentence or word expressed by a character string or a symbol string by an input means such as an OCR or a keyboard (not shown) is input. The input sentence or word is converted by the morphological analysis unit 8 into a morpheme string based on the contents stored in the morphological dictionary unit 9. The morphological dictionary unit 9 stores at least “reading” and “part of speech”. For example, “Kana-Kanji conversion by computer” by Aizawa and Ehara, NHH Technical Research, 25-5 Is divided into morphemes by means such as the longest match method described in (1). This longest matching method is an example of a means for morpheme division, and is not limited to this.

The morpheme analyzed by the morphological analysis unit 8 is supplied to a pitch control processing unit 10, and a pitch frequency component representing a pitch of a sound is determined based on the contents stored in the accent dictionary unit 11 and the accent combination rule unit 12. Is done.
The morpheme sequence to which the pitch frequency component is added is provided to the voice parameter generation unit 13 and is converted into a sequence of phoneme parameters based on the contents stored in the phoneme segment dictionary unit 14. The phoneme segment dictionary unit 14 holds a phoneme segment, that is, a phoneme constituting a sentence or a word, or a phoneme chain such as CV or VC as a parameter, and stores a phoneme or a phoneme constituting a morpheme.
The phoneme segments are arranged in the order of phonemes such as CV and VC, and the parameters in the unit frame are repeatedly converted into a speech parameter sequence at the pitch frequency intervals described above.

Hereinafter, the spectrum analysis unit 15, the spectrum change analysis unit 16,
The synthesizing unit 17, the utterance speed control unit 18 and the D / A conversion unit 19 perform the same operations as in the embodiment shown in FIG.

[Effects of the Invention] As described above, according to the present invention, when changing the utterance speed of synthesized speech, a temporal change rate is obtained by using a spectrum parameter, and the characteristic is determined based on the temporal change rate. Since the utterance speed is controlled by determining the synthesis frame period of the parameters, and the time of the synthesized speech signal is expanded or contracted based on the determined synthesis frame period, it is possible to prevent collapse of the phoneme whose spectrum changes drastically due to time expansion and contraction. And a natural and easy-to-listen sound can be obtained.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of one embodiment of the present invention.
FIG. 2 (a) is a diagram showing an example of a spectrum change rate, and FIG. 2 (b) shows the spectrum change rate shown in FIG. 2 (a) averaged for each frame. FIG. 2 (c) is an arrangement diagram of sound source pulses and spectrum parameters at the same speed as the original voice, and FIG. 2 (d) is based on the spectrum change rate shown in FIG. 2 (b). It is a figure which shows the arrangement | positioning of the determined new sound source pulse and the spectrum parameter. FIG. 3 is a schematic block diagram of another embodiment of the present invention. In the figure, 1 is an A / D converter, 2 is a sound source analyzer, 3 and 15 are spectrum analyzers, 4 and 16 are spectrum change analyzers, and 5 and 17
Is a synthesis unit, 6 and 18 are voice speed control units, 7 and 19 are D / A conversion units,
Reference numeral 8 denotes a morphological analysis unit, 9 denotes a morphological dictionary unit, 10 denotes a pitch control processing unit, 11 denotes an accent dictionary unit, 12 denotes an accent combination rule unit, 13 denotes a speech parameter generation unit, and 14 denotes a phoneme segment dictionary unit.

 ────────────────────────────────────────────────── ─── Continuing from the front page (72) Inventor Yoichi Higashikura 5th Sanraya, Daiya, Seika-cho, Soraku-gun, Kyoto Pref. ATI Co., Ltd. Vision Auditory Research Institute (56) References JP-A 62-245298 (JP, A) JP-A-63-64098 (JP, A)

Claims (2)

(57) [Claims] 1. An input means for inputting voice, A / D conversion means for converting voice input from the input means into a digital signal, and a digital voice converted by the A / D conversion means. Analysis means for analyzing a signal to extract feature parameters; synthetic speech conversion means for converting the feature parameters extracted by the analysis means into synthesized speech; and a synthesized speech converted by the synthesized speech conversion means into an analog signal. A voice synthesizing device comprising: a D / A converter that converts and outputs the digital voice signal; a spectrum analyzer that converts the digital voice signal converted by the A / D converter into a spectrum parameter that carries phonological information; Spectrum change analysis means for determining a temporal change rate using the spectrum parameter converted by the analysis means; and Based on the temporal rate of change determined by the spectrum change analysis means, comprising a speech rate control means for controlling a speech rate by determining a synthesis frame period of the feature parameter, wherein the speech synthesis means comprises: An utterance speed control circuit for a speech synthesizer, which performs time expansion and contraction of a synthesized speech signal based on a synthesized frame period determined by a speed control means. 2. An input means for inputting a character string or a symbol string; a reading string conversion means for converting a character string or a symbol string input by the input means into a reading string; A speech synthesis device comprising: a synthesized speech conversion unit configured to convert a read string converted by the synthesis speech into a synthesized speech; and a D / A conversion unit configured to convert a synthesized speech converted by the synthesized speech conversion unit into an analog signal. Spectrum analyzing means for converting the read string converted by the read string converting means into spectral parameters bearing phonological information; spectral change analyzing means for obtaining a temporal change rate using the spectral parameters converted by the spectral analyzing means And, based on the temporal change rate obtained by the spectrum change analysis means, An utterance speed control unit that determines a synthetic frame period and controls an utterance speed, wherein the synthesized voice conversion unit performs time expansion and contraction of the synthesized voice signal based on the synthesized frame period determined by the utterance speed control unit An utterance speed control circuit for a speech synthesizer, characterized in that: