JPH01120599A - Voice synthesization system - Google Patents

Abstract

PURPOSE: To improve the naturality of a synthesized voice having a long vowel to improve the tone quality by extending the vowl part of a syllable just prior to the long vowel to use it as a long vowel part at the time of synthesis of the long vowel. CONSTITUTION: When an input character string to be subjected to voice synthesis is inputted, a syllable read part 4 reads out voice parameters of required syllables from a syllable storage part 5 and sends them to a syllable coupling part 6. Mean-while, a time length setting part 7 sets the time length of each syllable in accordance with the input character string. A long vowel detection part 10 monitors the input character string; and when a long vowel is detected, the detection part 10 informs the syllable coupling part 6 that the vowel of the syllable just prior to this long vowel should be extended. The coupling part 6 couples voice parameters in accordance with time lengths of syllables set by the setting part 7. If read parameters and set time lengths are not consistent at thie time, time lengths of vowel parts are extended/reduced to perform adjustment. Vowels related to long vowel parts are extended simultaneously, and coupled voice parameters are sent to a waveform synthesis part 9, thus synthesizing a natural voice.

Description

[Detailed Description of the Invention] [Overview] The present invention analyzes the syllable-based sounds uttered by humans at short time intervals, and stores this as parameter time-series data for each syllable. Regarding a speech synthesis method that synthesizes arbitrary speech by combining speech composed of these parameter time series data, we aim to obtain synthesized speech with high naturalness and intelligibility when the speech to be synthesized includes long sounds. The purpose of this method is to provide a means for synthesizing long sounds by stretching the CI: syllable part of the syllable immediately before the long sound and using it as the long sound part.

[Industrial Application Field] Among the speech synthesis methods, the present invention analyzes syllable-based speech actually uttered by humans at short time intervals, and uses this as parameter time-series data of the speech for each syllable. This method relates to a speech synthesis method that synthesizes arbitrary speech by combining speech composed of these parameter time-series data stored in Regarding the method.

[Prior Art] Methods for artificially synthesizing speech can be roughly divided into three methods: ■ Recording/editing method, ■ Analysis/synthesis method, and ■ Pure synthesis method.

Of these, the recording/editing method (■) combines pre-recorded human voice waveforms and synthesizes them, and has the advantage of relatively simple equipment and control, and good sound quality, but the amount of information is very large. Since there are a lot of data, there is a drawback that a large capacity storage device is required.

In addition, the pure synthesis method (■) synthesizes speech from characters, etc. using certain synthesis rules, and there is no need to analyze human speech in advance. , the above-mentioned synthesis rule becomes very complicated, and its implementation is not easy.

On the other hand, the analysis-synthesis method (2) analyzes human speech in advance, converts the speech into parameters representing its characteristics, stores them, and synthesizes speech based on these parameters. In this method, instead of converting the audio into waveforms, the audio is converted into parameters that represent the shape of the subectals resulting from its articulation and parameters that represent the state of the sound source, and these information is compressed and stored. In recent years, speech synthesis methods using this method have been losing ground, as it requires less storage capacity and provides relatively good sound quality.

Based on this method, human utterances are analyzed in syllable units (for example, "a", "i", "tsu", etc.), and these syllables are combined to determine the pitch of the voice. A method of synthesizing arbitrary speech by controlling M is widely used.

[Problems to be Solved by the Invention] In the speech synthesis method for arbitrary words based on the conventional analysis and synthesis method as described above, long sounds are synthesized in the same manner as vowel chains of the same type.

For example, "Key" and "Kii" (strange) were synthesized in exactly the same way by connecting the syllable "i" after the syllable "ki" (homogeneous vowel chain).

Not only is the sound of the long note unnatural,
There was a problem in that it was not possible to distinguish between long sounds and homologous vowel chains. .

In view of these conventional problems, it is an object of the present invention to provide a speech synthesis method for arbitrary words that allows the pronunciation of long sounds to be natural and that also makes it possible to clearly distinguish between long sounds and homologous vowel chains. There is.

[Means for Solving the Problems] According to the present invention, the above objects are achieved by the means described in the claims. That is, the present invention analyzes syllable-based sounds uttered by humans at short time intervals, and uses this as parameter time-series data of the sounds.
In addition to the speech synthesis method that synthesizes arbitrary speech by combining the sounds made up of these parameter time series data that are accumulated for each syllable, when synthesizing a long sound, the vowel of the syllable immediately before the long sound is used. Speech synthesis method that expands the part and uses it as a long part [ ] Figure 1 is a diagram explaining the present invention in detail, (a) shows "unusual" speech synthesis, and (b) shows " 1 is the syllable “ki”.
, 2 represents the syllable "i", and 3 represents the extended part.

Conventionally, both "oddness" and "key" were determined by the method shown in (a), but according to the present invention, in the case of "keeper", (b)
By performing speech synthesis using the method shown in (a), it is possible to obtain synthesized speech that clearly distinguishes between such long sounds and homogeneous vowel chains as in case (a).

[Embodiment] FIG. 2 is a functional block diagram of an embodiment of the present invention,
4 is a syllable reading part, 5 is a syllable storing part, 6 is a syllable combining part,
Reference numeral 7 represents a time length setting section, 8 a pitch pattern setting section, 9 a waveform synthesis section, and 10 a long tone detection section.

In the figure, when an input character string to be synthesized into speech is input, a syllable reading section 4 reads out the speech parameters of the necessary syllables from a syllable storage section 5 and sends them to a syllable combining section 6.

On the other hand, a time length setting unit 7 sets the time length of each syllable from the input character string.A long sound detection unit 10 monitors the input character string, and when a long sound is detected, the syllable immediately before it. The syllable combining unit 6 is contacted to stretch the vowel.

The syllable combining unit 6 combines voice parameters according to the syllable time length set by the time length setting unit 7.

At this time, if there is a discrepancy between the read parameters and the set time length, adjustments are made by expanding or contracting the vowel part time length. At the same time, the vowel expansion related to the long part is performed, and the combined voice parameters are converted into a waveform. Send it to the synthesis section 9.

The waveform synthesis section 9 synthesizes a speech waveform using the speech parameters and the pitch pattern set by the pitch pattern setting section 8.

[Effect of the Invention 1] As described above, the present invention has the advantage of increasing the naturalness of synthesized sounds having long sounds and improving the sound quality.

It also has the effect of making it clearer to distinguish between long sounds (for example, “a-ichi”), which could not be distinguished in the past, and the same vowel chain (“aa”) or the same vowel chain.

[Brief explanation of the drawing]

FIG. 1 is a diagram explaining the present invention in detail, and FIG. 2 is a functional block diagram of an embodiment of the present invention. 1... syllable "ki", 2... syllable "i"
, 3...Extension unit, 4...Syllable reading unit,
5...Syllable storage section, 6...Syllable combining section, 7...Time length setting section, 8...Pitch pattern setting section, 9... ...Waveform synthesis section, 10...
...to long sound detection section

Claims (1)

[Claims] The syllable-based sounds uttered by humans are analyzed at short time intervals, and this is stored as parameter time-series data for each syllable, and these parameter time-series data are A speech synthesis method that synthesizes arbitrary speech by combining sounds consisting of , and is characterized in that when synthesizing a long sound, the vowel part of the syllable immediately before the long sound is stretched and used as the long sound part. method.