JPH0738113B2 - Speech synthesizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] The present invention analyzes a syllable-based voice uttered by a human at short time intervals, and stores it as parameter time series data of the voice for each syllable. , By combining the speech consisting of these parameter time series data,
Regarding a speech synthesis method for synthesizing arbitrary speech, the purpose is to provide a means for obtaining a synthesized speech with high naturalness and intelligibility when the speech to be synthesized includes a long sound. It is constructed by expanding the vowel part of the immediately preceding syllable and using it as a long sound part.

INDUSTRIAL APPLICABILITY The present invention analyzes voices in a syllable unit actually uttered by a human at short time intervals in a voice synthesizing method, and uses this as parameter time series data of the voices for each syllable. The present invention relates to a speech synthesis method for synthesizing arbitrary speech by accumulating speeches composed of these parameter time-series data, and particularly, speech synthesis that can improve the naturalness of speech when accompanied by long sound. It depends on the method.

[Prior Art] The methods of artificially synthesizing speech are roughly divided into three: recording and editing, analysis and synthesis, and pure synthesis.
It is divided into methods.

Among these, the recording / editing method is one in which prerecorded human voice waveforms are connected and synthesized, and the device and control are relatively simple and the sound quality is good, but the amount of information is very large. Therefore, there is a drawback that a large-capacity storage device is required.

In addition, the pure synthesis method is a method of synthesizing voices from characters etc. using a certain synthesis rule, and it is not necessary to analyze human voices in advance, but in order to create a voice with good naturalness, The composition rule becomes very complicated, and its implementation is not easy.

On the other hand, the analysis and synthesis method of is a method of analyzing a person's voice in advance, converting the voice into a parameter representing the characteristic and storing it, and synthesizing the voice based on this. In this method, the voice is not a waveform, but
Since the voice can be converted into a parameter indicating the shape of the spectrum by the articulation and a parameter indicating the state of the sound source and the information can be compressed and stored, the storage capacity is small and the sound quality is relatively high. In recent years, a voice synthesis method using this method has become widespread because a good one can be obtained.

Based on such a method, human uttered voices in syllable units (for example, “A”, “I”, “U”, etc.) are analyzed and placed, and the syllables are combined to control the pitch of the voice. Therefore, a method of synthesizing an arbitrary voice 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 by the same method as the vowel chain of the same kind.

For example, both "key" and "key" (odd) were synthesized by connecting the syllable "i" to the syllable "a" (same vowel chain) in exactly the same way.

Therefore, not only is the sound of the long tones unnatural,
There was a problem that it was not possible to distinguish between long consonants and vowel chains of the same kind.

In view of such conventional problems, the present invention has an object to provide a speech synthesis method for an arbitrary word, which is natural in voicing long sounds and which can clearly distinguish between long sounds and consonant vowel chains. There is.

[Means for Solving the Problems] According to the present invention, the above object is achieved by the means described in the claims. That is, the present invention analyzes voiced syllable units at short time intervals, accumulates this as parameter time-series data of the voice for each syllable, and extracts from these parameter time-series data. In a speech synthesizer for synthesizing an arbitrary speech by combining the following speeches, the speech synthesizer uses the vowel part of the syllable immediately before the long sound as a long sound portion when the long sound is synthesized.

[Operation] FIG. 1 is a diagram for explaining the principle of the present invention, in which (a) shows "odd" voice synthesis and (b) shows "key" voice synthesis. Reference numeral 1 represents a syllable “ki”, 2 represents a syllable “a”, and 3 represents an expanded portion.

Conventionally, both the "odd" and the "key" are synthesized by the method shown in (a), but according to the present invention, by performing the speech synthesis by the method shown in (b) in the case of the "key", It is possible to obtain a synthetic voice in which the distinction between such a long sound and the homologous vowel chain as in the case of (a) can be clearly heard.

[Embodiment] FIG. 2 is a functional block diagram of an embodiment of the present invention.
4 is a syllable reading unit, 5 is a syllable storage unit, 6 is a syllable connecting unit, and 7
Is a time length setting unit, 8 is a pitch pattern setting unit, 9 is a waveform synthesizing unit, and 10 is a long sound detecting unit.

In the figure, when an input character string to be voice-synthesized is input, the voice reading unit 4 reads a voice parameter of a required syllable from the syllable storage unit 5 and sends it to the syllable combining unit 6.

On the other hand, the time length setting unit 7 sets the time length of each syllable from the input character string.

Further, the long sound detecting unit 10 monitors the input character string, and when a long sound is detected, the long sound detecting unit 10 informs the syllable connecting unit 6 to extend the vowel of the syllable immediately before that.

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

At this time, if there is a disagreement between the read parameter and the set time length, the vowel part time length is expanded or contracted for adjustment. At the same time, vowel expansion related to the long sound portion is performed, and the combined voice parameters are sent to the waveform synthesizing unit 9.

The waveform synthesizing section 9 synthesizes a speech waveform with the speech parameter and the pitch pattern set by the pitch pattern setting section 8.

[Effects of the Invention] As described above, according to the present invention, there is an advantage that the naturalness of a synthetic sound having a long sound can be increased and the sound quality can be improved.

Then, there is an effect that distinction between a long sound (for example, “Ah”) and the same vowel chain (“Aa”) or the same kind of vowel chain, which cannot be distinguished conventionally, becomes clear.

[Brief description of drawings]

FIG. 1 is a diagram for explaining the principle of the present invention, and FIG. 2 is a functional block diagram of an embodiment of the present invention. 1 ... syllable "ki", 2 ... syllable "a", 3 ... extension part,
4 ... syllable reading unit, 5 ... syllable storage unit, 6 ... syllable combining unit, 7 ... time length setting unit, 8 ... pitch pattern setting unit, 9 ... waveform synthesis unit, 10 ... long sound detecting unit

Claims (1)

[Claims] 1. Spoken syllable-based speech is analyzed for each short time interval, and this is stored as parameter time-series data of the speech for each syllable, and is composed of these parameter time-series data. By combining the voice,
A voice synthesizing device for synthesizing an arbitrary voice, wherein when synthesizing a long sound, a vowel part of a syllable immediately before a long sound is expanded and used as a long sound part.