JPH09152884A - Speech synthesizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To learn the speech pattern of a user and synthesize a natural voice by providing a speech recognition part and a feature parameter storage part which learns speech information and updates feature parameters. SOLUTION: A phoneme dictionary part 3 of a feature parameter storage part which learns speech information from the speech recognition part 7 and updates feature parameters of a voice to be synthesized is provided with a phoneme dictionary update part 3A as a RAM which stores phonemes that a user generates from a microphone 6. Further, a rhythm rule part 4 is provided with a rhythm rule update part 4 which stores the intonation and accent of the voice that the user generates from the microphone 6 through a filter bank 8. Then the voice spoken for speech recognition is inputted to the speech recognition part 7 to analyze the feature parameters of the voice, which are updated when a mode wherein the parameter update is performed is entered, thereby synthesizing the voice by using the feature parameters of the user.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a regular speech synthesizer for converting a sequence represented by discrete symbols such as letters and phonetic symbols into continuous speech, and more particularly to learning a user's speech pattern to synthesize a natural speech. The present invention relates to a speech synthesizer that enables

[0002]

2. Description of the Related Art FIG. 7 is a diagram showing an example of a conventional speech synthesizer. The same components are denoted by the same reference numerals or symbols throughout the drawings. As shown in this figure,
The voice synthesizing device is stored in the vehicle for providing the voice synthesizing unit 2 with a voice synthesizing unit 2 for synthesizing information on characters and symbol strings from a navigation unit 1 for guiding a road, which is installed in a vehicle, into a voice. A phoneme dictionary unit 3 formed of a ROM (Read Only Memory), a prosody rule unit 4 formed of a ROM for storing a voice intonation, accent, etc. for giving to the voice synthesis unit 2, and a voice synthesized by the voice synthesis unit 2. And a speaker 5 for outputting. The phoneme dictionary unit 3 and the prosody rule unit 4 constitute a characteristic parameter storage unit that stores characteristic parameters.

The phoneme dictionary unit 3 stores phonemes such as vowels and consonants. FIG. 8 is a diagram showing intonations and accents stored in the prosody rule unit 4. As shown in the figure, the prosody rule section 4 stores intonation in which the pitch frequency decreases with time and accents in which the time change of the pitch frequency has a trapezoidal shape.

The prosody rule section 4 has an intonation,
Several types of accents are provided, and one can be selected according to the user's preference.

[0005]

However, in the above speech synthesizer, it is possible to select the intonation and accent according to the user's preference, but since the selection range is limited, the user can select the intonation and accent. There is a problem that the degree of freedom is small and the intonation and accent do not always match the user's preference.
Further, since phonemes are also given unilaterally, there is a problem that the phonemes are not suited to the user's preference.

Therefore, in view of the above problems, it is an object of the present invention to provide a speech synthesizer capable of giving intonation, accents, and phonemes according to user's preference.

[0007]

In order to solve the above problems, the present invention provides a speech synthesizer having the following configuration. That is, a voice synthesizing device for synthesizing characters and symbol strings into speech, a system that responds to the voice information with the characters and symbol strings, and a voice recognition unit that recognizes the voice information
A feature parameter storage unit having a feature parameter for synthesizing the voice, and learning the voice information to update the feature parameter is provided. With this device,
By learning the utterance pattern of the user, it becomes possible to select the characteristic parameter according to the user's preference, which enables natural speech synthesis.

In the characteristic parameter storage section, the prosody information of intonation and accent extracted from the voice information is updated. By this means, it is possible to learn the prosody information of the user's utterance and generate a synthetic sound having a prosody according to the user's language environment. The intonation and accent extracted in the characteristic parameter storage section are separately used for synthesis. By this means, similarly to the above, it is possible to learn the prosody information of the user's utterance and generate a synthetic sound having a prosody according to the language environment of the user.

Specifically, the frequency analysis of the voice information is performed, and the intonation and accent prosody to be updated are updated based on the time change of the pitch frequency. The feature parameter storage unit updates the phoneme dictionary data to the phonemes extracted from the voice information. By allowing the user's voice to be registered by this means, the user's own voice can be synthesized, and a synthesized voice having a prosody according to the user's language environment can be generated.

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a speech synthesizer according to the present invention. In contrast to the configuration of FIG. 7, the voice synthesizer includes a microphone 8, a voice recognition unit 7 connected to the microphone 6 for outputting a voice recognition result to the navigation 1, and a plurality of band pass filter groups connected to the microphone 6. Further, a filter bank 8 for analyzing the false peach frequency is additionally provided. Further, in order to learn the voice information from the voice recognition unit 7 and update the feature parameter of the voice to be synthesized, the phoneme dictionary unit 3 of the feature parameter storage unit stores the phonemes generated by the user from the microphone 6. RAM (Random Access Memory) for
Has a phoneme dictionary updating unit 3A, and the prosody rule unit 4 has an intonation of a voice generated by a user from a microphone 6 via a filter bank.
A prosody rule updating unit 4A is provided as a RAM for storing accents.

When the user utters "I want you to find a nearby gas station" into the microphone 6, the voice recognition device 7 recognizes this voice and the navigation 1
To order a gas station search. From the navigation 1, as a result of the search, the speech synthesizer 2 synthesizes a symbol string "There is a gas station of XX in 100 meters ahead", and it is output from the speaker 5.

The navigation 1 is an example, and another system may be used as long as it responds to the voice information from the voice recognition section 7 with the character and symbol strings. FIG.
Is the pitch period 1 / analyzed in the filter bank 8 of FIG.
It is a figure which shows f. As shown in this figure, the filter bank 8
Then, using the user's utterance to the voice recognition device, the time change of the pitch frequency is measured, the intonation and the accent are obtained, and stored in the phoneme dictionary updating unit 3A. Also,
This result is also used as the user's speaking speed (number of mora / second) and voice pitch.

In the prosody rule updating unit 4A, the user specially utters dozens of types of phonemes, and the result is stored. In this way, using the voice recognition device,
Instead of uniform voice synthesis, the intonation and accent of the user's utterance are synthesized, so that it is possible to generate a synthetic sound having a prosody according to the user's language environment. Furthermore, by registering the user uttered phonemes, it becomes possible to synthesize by the user's own utterance.

FIG. 3 is a first flowchart showing a series of basic operations of the speech synthesizer of the present invention. Step S1
At, voice input to the voice recognition unit 7 is performed. In step S2, it is determined whether the parameter update is completed.
The user determines parameter updating and selection. The parameters include a plurality of intonations, accents, vocalization rates, pitches, and the like.

In step S3, characteristic parameters of the input voice are extracted. In step S4, the characteristic parameter of the synthetic sound is updated, and the process returns to the completed step S2. In step S5, voice recognition processing is performed. In step S6, navigation processing is performed.

In step S7, a voice synthesis process is performed on the navigation command using the updated parameters. Therefore, according to the present invention, the voice uttered for voice recognition is input to the voice recognition unit 7, the characteristic parameter of the voice is analyzed, and at this time, the mode is set to update the parameter. If
The characteristic parameter of the voice is updated, and the voice synthesis is performed using the characteristic parameter of the user. That is, the user's utterance pattern is learned to enable more natural voice synthesis.

FIG. 4 is a second flowchart showing a modification of FIG. Steps S1, 2, 5, 6, 7 are the same. In step S8, prosody information of the input voice is extracted. In step S9, the intonation rule is updated. In step S10, the accent rule is updated.

Therefore, according to the present invention, the voice uttered for voice recognition is input to the voice recognition unit 7 and the prosody information of this voice is extracted. At this time,
If the mode for updating parameters is set, the prosody rule unit 4 such as the intonation rule and accent rule of the voice is updated, and voice synthesis is performed using the updated parameters. That is, by learning the prosody information of the user's utterance, it is possible to generate a synthesized voice having a prosody according to the user's language environment.

FIG. 5 is a third flowchart showing a modification of FIG. Steps S1, 2, 5, 6, 7, 8 are the same. In step S11, it is determined whether there is an intonation update. If this determination is "YES", the process proceeds to step S12, and if "NO", the process is step S13.
Proceed to.

In step S12, the intonation rule is updated. In step S13, it is determined whether or not there is an accent update. This judgment is "Y
If "ES", the process proceeds to step S14, and if "NO", the process returns to step S2. In step S14, the accent
After updating the prosody rule part 4 such as rules, step S2
Return to

Therefore, according to the present invention, the voice uttered for voice recognition is input to the voice recognition unit 7 and the prosody information of this voice is extracted. At this time,
If the intonation and the accent are in the update mode, the prosody rule section 4 such as the intonation rule and the accent rule are updated, and the voice synthesis is performed using the updated parameters. That is, by learning the prosody information of the user's utterance, it is possible to generate a synthesized voice having a prosody according to the user's language environment.

FIG. 6 is a fourth flowchart showing updating of phoneme data. Steps S1, 5, 6, 7 are the same. In step S15, it is determined whether phoneme registration has ended. If this determination is "YES", the process proceeds to step S5, and if "NO", the process proceeds to step S16. Step S1
6, the phoneme of the user is registered, and step S1
Return to 5.

Therefore, according to the present invention, the voice uttered for voice recognition is input to the voice recognition unit 7, and the phoneme dictionary 3 is updated if the user voice registration mode is set. , Synthesize with user's voice. That is, by allowing the user's voice to be registered, it is possible to synthesize the voice with the user's own voice.

[Brief description of the drawings]

FIG. 1 is a diagram showing a speech synthesizer according to the present invention.

2 is a diagram showing a pitch period 1 / f analyzed by the filter bank 8 of FIG. 1. FIG.

FIG. 3 is a first flowchart showing a series of basic operations of the speech synthesizer of the present invention.

FIG. 4 is a second flowchart showing a modified example of FIG.

5 is a third flowchart showing a modified example of FIG.

FIG. 6 is a fourth flowchart showing updating of phoneme data.

FIG. 7 is a diagram showing an example of a conventional speech synthesizer.

8 is a diagram showing intonation and accents stored in the prosody rule unit 4. FIG.

[Explanation of symbols]

 1 ... Navigation 2 ... Speech synthesis section 3 ... Phoneme dictionary section 3A ... Phoneme dictionary update section 4 ... Prosodic rule section 4A ... Prosodic rule update section 5 ... Speaker 6 ... Microphone 7 ... Speech recognition section 8 ... Filter bank

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Ikue Takahashi 1-2-2 Goshodori, Hyogo-ku, Kobe-shi, Hyogo Within Fujitsu Ten Limited

Claims (5)

[Claims] 1. A voice synthesizing apparatus for synthesizing characters and symbol strings into speech, comprising: a system for responding to speech information with the characters and symbol strings; a speech recognition unit for recognizing the speech information; and synthesizing the speech. And a feature parameter storage unit that learns the voice information and updates the feature parameter. 2. The speech synthesizer according to claim 1, wherein the characteristic parameter storage unit updates the prosodic information of intonation and accent extracted from the speech information. 3. The speech synthesizer according to claim 2, wherein the intonation and the accent extracted in the characteristic parameter storage section are separately used for synthesis. 4. The speech synthesizer according to claim 2, wherein the intonation and accent prosody to be updated are updated based on the time change of the pitch frequency by performing frequency analysis of the voice information. 5. The speech synthesizer according to claim 1, wherein the characteristic parameter storage unit updates the phoneme data extracted from the speech information.