JP2847699B2 - Speech synthesizer - Google Patents

Description

Description: TECHNICAL FIELD [0001] The present invention relates to a voice synthesizing device that converts arbitrary message content input from a keyboard or the like into voice. [Prior Art] Conventionally, this type of speech synthesizer edits the parameters of CV syllables according to the input message content, generates a sound source waveform according to the accent and intonation of the speech message, and performs speech synthesis. A synthesized voice was obtained using a filter. However, simply connecting CV syllables causes the spectrum of the resulting synthesized speech to change rapidly at the connection points, so the synthesized speech is not only very discontinuous but also heard as an incorrect sound. Quite often, very poor quality. Some devices have an interpolation circuit that interpolates parameters so as not to cause abrupt changes.However, in this device, even if spectral continuity can be obtained by parameter interpolation, natural sound In many cases, a spectrum that cannot be seen is generated, and even with such a device, the sound may be heard as an erroneous sound, and there is a disadvantage that a high-quality synthesized speech cannot be obtained. Also, a device has been proposed in which the parameters of VCV syllables are stored, and the parameters are combined between the same vowels to obtain a high-quality synthesized sound. The number is increased by a factor of six or more, and there is a drawback in that there are many problems in terms of the capacity of the storage circuit and the time required for creating parameters. [Summary of the Invention] The present invention has been made for the purpose of improving the above-mentioned drawbacks of the conventional art. A storage circuit is provided to read CV syllable parameters and interpolation parameters according to the content of the message to be uttered from this storage circuit, and insert interpolation parameters into the read CV syllable parameters as appropriate by the interpolation circuit. Accordingly, the present invention provides a speech synthesizer capable of obtaining high-quality synthesized speech having natural continuity. Embodiments of the Invention Hereinafter, embodiments of the present invention will be described with reference to the drawings. First
FIG. 1 is a block diagram showing a speech synthesizer according to an embodiment of the present invention. In the figure, reference numeral 1 denotes a message content input from a keyboard or the like (not shown), 2 denotes a sound source waveform generating circuit for generating a sound source waveform having a pitch corresponding to the accent and intonation of a voice from the message content 1, and 3 denotes a sound source waveform generation Sound source waveform obtained from circuit 2, 4
Is a storage circuit that stores CV syllable parameters and interpolation parameters, 5 is a readout circuit that reads out necessary parameters according to message content 1, 6 is the readout CV syllable parameters and interpolation parameters, and 7 is interpolation Circuit, 8
Is an edited synthesis parameter, 9 is a voice synthesis filter circuit, and 10 is a synthesized voice. FIG. 2 is an explanatory diagram showing an example in which parameters are read and interpolated in the speech synthesizer shown in FIG. Next, the operation of the voice synthesizing apparatus according to the embodiment of the present invention shown in FIG. 1 will be described. The readout circuit 5 reads out parameters necessary for synthesis from the storage circuit 4 in accordance with the message content 1 input from a keyboard or the like (not shown). Now, in the case of synthesizing the word speech “future” (mi rai) as shown in FIG.
Is "mi" stored as a parameter of CV syllable,
The parameters “ra” and “i” are read. here,
When combining the “mi” and “ra” parameters, as shown by the hatched lines in FIG. 2, read the parameters of the rising part of the vowel from the parameters stored as the CV syllable parameter “ri”. The circuit 5 reads and inserts between each parameter "mi" and "ra" as shown by the arrow in the figure.
Also, in the combination of each parameter “ra” and “i”,
The readout circuit 5 reads out the parameter of “ai” stored as the interpolation parameter extracted from the natural voice “Ai” and inserts it as shown by the arrow in the figure. These parameters are interpolated by an interpolation circuit 7 in an interpolation section as shown by horizontal stripes in FIG. 2, that is, before and after the inserted parameter. A synthesized speech 10 is obtained by a speech synthesis filter circuit 9 from the synthesized parameters 8 thus obtained and the sound source waveform 3 generated by the sound source waveform generation circuit 2. This is based on the fact that in natural speech, the spectrum of a vowel changes greatly as it approaches a subsequent consonant, but this change appears in the opposite form in the combination of the same consonant and vowel. That is, in the present invention, the state of articulation coupling from a vowel to a consonant is
It utilizes the characteristic that the time axis of articulation coupling from a consonant to a vowel is close to the appearance when viewed in reverse.For this purpose, a synthesized sound is produced by inserting interpolation parameters and interpolating interpolation sections before and after the interpolation parameters. The quality is greatly improved. Also, since a part of the parameters of the CV syllables stored in advance are cut out and used, the parameters that need to be newly stored are CV syllables.
Parameters at the time of change from vowels to vowels not in syllable parameters and parameters at the time of change from vowels to semi-vowels, etc.
Very little. As described above, in the present invention, it is only necessary to prepare parameters for interpolation from a vowel that does not appear in an excessive section of a CV syllable to another vowel, and thereafter, data is read from the CV syllable and interpolation is performed. It is. [Effects of the Invention] As described above, the present invention provides, in a speech synthesizer, a storage circuit that stores a parameter of a CV syllable and an interpolation parameter to be inserted when combining the parameter of the CV syllable. The CV syllable parameters and interpolation parameters are read out from the circuit according to the content of the message to be uttered, and the interpolation circuit appropriately inserts the interpolation parameters into the read out CV syllable parameters. Thus, it is possible to eliminate the deterioration of the quality of the synthesized voice at the connection point of and to obtain an excellent effect that a high-quality synthesized voice having natural continuity can be obtained.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing a speech synthesizer according to an embodiment of the present invention, and FIG. 2 is an example in which parameters are read and interpolated in the speech synthesizer shown in FIG. FIG. In the figure, 1 ... message contents, 2 ... sound source waveform generation circuit, 3 ... sound source waveform, 4 ... storage circuit, 5 ... readout circuit,
6: CV syllable parameters and interpolation parameters, 7: interpolation circuit, 8: synthesis parameter, 9: speech synthesis filter circuit, 10: synthesized speech.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G10L 5/00-5/04

Claims (1)

(57) [Claims] In a speech synthesizer for storing feature parameters of CV syllables, editing the feature parameters according to the content of a message to be uttered, generating a sound source waveform, and synthesizing speech by a speech synthesis filter, A storage circuit that stores the feature parameters and stores an interpolation parameter that does not appear in the CV syllable, and according to the message content to be uttered from the recording circuit,
A readout circuit that reads out the feature parameter of the CV syllable, and reads out the feature parameter or the interpolation parameter of the CV syllable in accordance with the excessive section of the CV syllable, and the readout feature parameter of the CV syllable. , The CV read according to the excessive section of the CV syllable
An interpolation circuit for interpolating by inserting an interpolation parameter generated from a syllable characteristic parameter or an interpolation parameter consisting of the interpolation parameter; a sound source waveform generation circuit for generating a sound source waveform according to the content of the message to be uttered; A speech synthesizer, comprising: an interpolation circuit; and a speech synthesis filter circuit that synthesizes speech from parameters obtained by the sound source waveform generation circuit and a sound source waveform.