JPH0279097A - Pitch control system - Google Patents

Abstract

PURPOSE:To generate a pitch pattern with high naturalness corresponding to an actual case by expanding or reducing the variation range of a fundamental frequency calculated from a pitch control rule according to information obtained from an input symbol sequence, a control signal which is supplied from outside etc. CONSTITUTION:A pitch control part 5 refers to the pitch control pattern recorded in a pitch parameter file 6 at need by using information on symbols indicating a phoneme and a pitch in the input symbol sequence to set pitch continuance length, a fundamental frequency pattern, an amplitude pattern, etc., and a mean fundamental frequency changing part 8 alters a fundamental frequency pattern set by the pitch control part 5. Then a dynamic range changing part 9 expands or reduces the variation range of the fundamental frequency calculated from the pitch control rule according to instruction codes in the input symbol sequence 3, other information obtained from the input symbol sequence 3, a control signal which is supplied from outside, etc. Consequently, the natural pitch pattern corresponding to an actual case is obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a prosody control method, and more particularly to a prosody control method for naturally adding prosody in speech synthesis.

In order to add natural prosody, prosodic control rules that control the fundamental frequency, amplitude, and phoneme duration length are essential in the 1-stroke speech synthesis. The fundamental frequency pattern that is judged to be the most natural varies depending on many factors, such as the content of the sentence being uttered, the purpose of use, the form of use, and the user's preferences. For this reason, a method of adjusting and using a natural fundamental frequency pattern according to the situation by changing the average fundamental frequency has been proposed, for example, in Japanese Patent Laid-Open No. 61-70.
It is known from Japanese Patent No. 597.

Depending on the type of sentence to be uttered, the average fundamental frequency may be used instead of the average fundamental frequency to add natural prosody during speech synthesis.
When it becomes necessary to change the range of change of the fundamental frequency, that is, the dynamic range,6 for example, a wide dynamic range becomes a hindrance when reading out stock market conditions. When reading novels etc., rich intonation cannot be achieved with a narrow dynamic range. Even in the same novel, the dynamic range required for written text and conversational text are different.

For this reason, the above-mentioned method of changing only the average fundamental frequency cannot produce a natural prosodic pattern depending on the occasion.

Purpose The present invention was made in view of the above-mentioned circumstances.
In particular, the prosodic control rules are calculated based on the information obtained from the instruction code embedded in the input symbol string and other input symbol strings, the control signal given to the regular speech synthesizer from the outside when executing speech synthesis, etc. This was done for the purpose of making it possible to expand or reduce the variation range of the fundamental frequency.

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention reads out a parameter sequence of a speech segment prepared in advance according to an input symbol string expressing phoneme/prosody, and uses a speech parameter combination rule to read out a parameter series of a speech segment that is prepared in advance. In a regular speech synthesizer that adds prosody, the input symbols are The fundamental frequency calculated based on the prosodic control rule is determined according to information obtained from the instruction code embedded in the string and other input symbol strings, a control signal given to the rule speech synthesizer from the outside when executing speech synthesis, etc. It is characterized by expanding or contracting the range of change. Hereinafter, the present invention will be explained based on examples.

FIG. 1 is a block diagram for explaining one embodiment of the present invention, in which 1 is a speech segment file, 2 is a phoneme generator, 3 is an input symbol string, 4 is a speech synthesizer, and 5 is a speech unit file. 6 is a prosody control section, 6 is a prosodic parameter file, 7 is an utterance speed change section, 8 is an average fundamental frequency change section, 9 is a dynamic range change section, and the input symbol string contains symbols representing phoneme and prosody, and The instruction codes for changing the control parameters are mixed. The phoneme generation unit 2 selects necessary speech segment parameters from the speech segment parameter file, mainly using information on symbols representing phonemes in the input symbol string 3, and smoothly combines the individual speech segment parameters. Then, the phonological parameters of the synthesized speech are generated. The prosody control unit 5 uses the information on the symbols representing the phoneme and prosody in the input symbol string to refer to the prosody control parameters recorded in the prosody parameter file 6 as necessary, and determines the duration of the prosody, the fundamental frequency, etc. Set the pattern, amplitude pattern, etc.

The prosodic pattern set here is an average value of prosodic patterns in various types of sentences.6 The speech speed changing section 7 changes the phoneme duration length set by the prosody control section 5. The changes are made such that the speech speed dictated by the instruction code in the input symbol string is achieved. The average fundamental frequency changing section 8 changes the fundamental frequency pattern set by the prosody control section 5. The changes are made such that the average fundamental frequency dictated by the instruction code in the input string is achieved. The dynamic range changing section 9 changes the fundamental frequency pattern set by the prosody control section 5. The changes are made such that the average fundamental frequency dictated by the instruction code in the input string is achieved.

FIG. 2 is a block diagram for explaining another embodiment of the present invention, in which 10 indicates a control signal and 11 indicates a control signal input section. In this embodiment, the input symbol string is a symbol string representing phoneme and prosody.

Further, the explanation regarding the phoneme generation section and the prosody control section is the same as that of the embodiment shown in FIG. The speech speed changing unit 7 changes the phoneme duration set by the prosody control unit 5. The changes are made such that the speech speed dictated by the control signal input by the control signal input section during execution is achieved.

The average fundamental frequency changing section 8 changes the fundamental frequency pattern set by the prosody control section 5. The changes are made in such a way that the average fundamental frequency dictated by the control signal input from the control signal input during execution is achieved. The dynamic range changing section 9 changes the fundamental frequency pattern set by the prosody control section 5. The changes are made in such a way that the dynamic range dictated by the control signal input from the control signal input section during execution is achieved.

Effects As is clear from the above explanation, according to the present invention, the information obtained from the instruction code embedded in the input symbol string and other input symbol strings, and the information given from the outside to the regular speech synthesizer during speech synthesis execution. By expanding or reducing the variation range of the fundamental frequency calculated from the prosody control rule in accordance with the control signal, a highly natural prosody pattern can be generated depending on the occasion.

[Brief explanation of the drawing]

FIG. 1 and FIG. 2 are configuration diagrams for explaining the present invention in detail, respectively. 1... Speech element file, 2... Phoneme generation unit, 3.
... input symbol string, 4 ... speech synthesizer, 5 ... prosody control section, 6 ... prosody parameter file, 7 ... speech speed change section, 8 ... average fundamental frequency change section, 9・
...dynamic range changing section, 10...control signal,
11... Control signal input section. Figure 1 Figure 2 Figure U11

Claims (1)

[Claims] 1. The parameter series of the speech segments prepared in advance is
It reads out according to the input symbol string expressing the prosody, connects the speech unit parameter series according to the speech parameter combination rule, and uses the prosody control rule to read out the phoneme duration length and basic according to the input symbol string expressing the phoneme and prosody. In a regular speech synthesizer that adds prosody by calculating frequency patterns, amplitude patterns, etc., information obtained from instruction codes embedded in input symbol strings and other input symbol strings, and rules from outside when executing speech synthesis are used. A prosody control method characterized by expanding or reducing a change range of the fundamental frequency calculated from the prosody control rule in accordance with a control signal given to a speech synthesizer.