JPS62143098A - Synthesization of monosyllable for evaluating clearness - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a method for synthesizing monosyllables used for evaluating synthesized speech synthesized according to rules.

(Prior Art) In general, in the rule synthesis of speech, monosyllables are synthesized in a rule to evaluate the intelligibility of the synthesized speech.

FIGS. 1 and 2 are syllable combination diagrams showing a conventional method for synthesizing monosyllables according to rules. Figure 1 is a monosyllable power diagram.
The interval between 1 and 2 is a silent interval, the interval between 2 and 3 is a transient interval, and the interval between 3 and 5 is a voiced interval. Note that some syllables do not have the above-mentioned voiceless sections Fin 1 or Fin 2.

Conventionally, the speech power in rule synthesis has ranged from above to 3.
For the unvoiced section and the transient section, the voice power obtained by spectrum analysis is used, and for the section 3 and 4, linear interpolation is performed, and for the straight section 4 and 5, a linear supplementary law is applied so that the voice power becomes 0 in frame 5. ing. Similarly, the spectral parameters used for spectrum analysis are used only for sections 1 to 3 and frame 4. By adjusting the lengths of 3 to 4 and 4 to 5, an arbitrary time length of the syllable can be obtained.

FIG. 2 shows a pitch pattern, which has conventionally been synthesized using pitch frequencies exhibiting flat characteristics as shown by 6.7 and 8.

However, such conventional monosyllabic rule synthesis
In order to match the syllable length to that of the original speech, each syllable was different. Furthermore, because the pitch pattern was flat, the synthesized sound had a musical-like sound, which was unnatural and difficult to hear.

Such synthesized sounds have the disadvantage that subjects in speech intelligibility tests cannot concentrate on identifying syllables.

(Problems to be Solved by the Invention) The present invention eliminates the above-mentioned conventional drawbacks, reduces the psychological discomfort of subjects in speech intelligibility tests, accurately measures syllable intelligibility, and makes correct judgments. The purpose is to make people do the following.

(Means for Solving the Problems) In order to achieve the above object, the present invention sets a standard syllable length to reduce variations in syllable length among a plurality of syllables in a speech intelligibility test. , also by giving a convex variation to the pitch pattern.

This eliminates the sound of musical tones and produces a regular synthesized sound that is close to natural speech.

(Function) By forming the above-mentioned rule-based synthesized speech, it becomes possible to reduce the psychological discomfort of the subjects in the speech intelligibility test, to accurately measure the intelligibility of syllables, and to make them make correct judgments.

(Example) Hereinafter, the present invention will be explained in detail by way of an example with reference to FIGS. 1 and 2 mentioned above.

First, in FIG. 2, the pitch patterns indicated by 6, 9, and 10 are pitch patterns for implementing the present invention, and the pitch pattern increases from the beginning of transition section 2 or 3 in FIG. 1 to vowel interpolation start frame 3. The pitch pattern is convex and descends from the vowel interpolation start frame 3 to the end frame of the syllable. Here, the pitch frequencies at positions 6.9 and 10 above are respectively 214
t(z, 222H2 and 20911z, and this value was obtained by averaging the pitch patterns extracted from natural speech. Similarly, the syllable length is 224 m.
5, and long syllables such as /Shu/ were handled as an exception, and monosyllables were formed by regular synthesis.

The table below shows the rules for synthesizing the pitch patterns with the same syllable length and pitch height, which do not sound like musical tones, obtained by varying the pitch pattern and determining the syllable length as described above. This figure shows the results of a speech intelligibility test using speech. As will be seen, the monosyllables for intelligibility evaluation according to the invention have improved syllable intelligibility.

(Effects of the Invention) As is clear from the above explanation, in the present invention, since the time lengths of each syllable are approximately equal, the speech intelligibility test subject does not need to be concerned about the difference in syllable time length. , you can concentrate on the syllable identification task, and since it does not have a musical sound, you can concentrate on the syllable identification task without feeling uncomfortable.Therefore, it has the advantage of being able to accurately evaluate speech intelligibility. A noticeable effect can be obtained.

[Brief explanation of drawings]

FIGS. 1 and 2 are syllable combination diagrams and pitch pattern diagrams for explaining a conventional example and the present invention. Work or 2 (section)... Silent section, 2 or 3 (
section)...Transient section, 3 to 4.4 to 5
...Linear interpolation interval.

Claims (1)

[Claims] In the rule synthesis of monosyllables used for the intelligibility test of rule-synthesized speech, without flattening the pitch pattern,
A method for synthesizing a single syllable for intelligibility evaluation, characterized in that the syllable is synthesized by approximating the pitch pattern of a syllable in natural speech, and by keeping the duration of the syllable almost constant regardless of the syllable.