JPH09230884A - Text voice conversion method and method for determining breathing position of text voice converter - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to convert inputted sentences to natural speeches by determining breathing positions in accordance with the inputted sentences and inserting breathing tones in the determined breathing positions. SOLUTION: The speech signals formed in a speech forming section 2 is sent to a breathing and silent inserting section 3 and are sent to a breathing position candidate detecting section 5. The breathing and silent inserting section 3 inserts the breathing or silent speech signals into the speech signals formed in the speech forming section 2 according to the controlling command from the breathing and silent inserting control section 6. The speech signals obtd. by the breathing and silent inserting section 3 are outputted as speeches via a speech output section 4. The breaching position candidate detecting section 5 accumulates the power values of the speeches from the speech signals formed by the speech forming section 2 and sends breaching position information to a breathing and silent insertion control section 6 when the cumulative value attains the prescribed value determined beforehand. The control section clears the cumulative value thus far and freshly starts the accumulation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sentence / speech conversion method and a breathing position determining method in a sentence / speech conversion apparatus.

[0002]

2. Description of the Related Art In a sentence-to-speech converter that generates a voice corresponding to an input sentence, it is important to generate a natural voice corresponding to the input sentence. In order to generate a natural voice corresponding to an input sentence, a pose is inserted in the generated voice. The insertion position of the pose is determined based on the punctuation marks of the sentence, the boundaries of the meaning, the elapsed time from the final breathing position, and the like. However, when it occurs in a loud voice, shortness of breath occurs, so that a physiological phenomenon such as an increase in breathing is not reflected.

[0003]

SUMMARY OF THE INVENTION It is an object of the present invention to provide a sentence-speech conversion method capable of converting an input sentence into a more natural voice and a breathing position determining method in a sentence-speech conversion device.

[0004]

A sentence-speech conversion method according to the present invention is a sentence-speech conversion method for converting an input sentence into a voice, in which a breathing position is determined based on the input sentence and the determined breathing position is determined. It is characterized by inserting a breathing sound at the position.

The breathing position is determined as follows, for example. (1) The breathing position is determined based on the cumulative value of the voice power of the voice signal generated from the input sentence.

(2) The break position in grammatical / semantic analysis first appearing after the breathing position candidate determined based on the cumulative value of the voice power of the voice signal generated from the input sentence is determined as the breathing position. To do.

In the sentence-speech conversion method according to the present invention, since the breathing sound is inserted at the breathing position instead of the silence,
A more natural sound can be obtained.

A first breath-speech position deciding method in the first sentence-speech conversion apparatus according to the present invention is a breath-speech position deciding method in a sentence-speech conversion apparatus for converting an inputted sentence into a voice, and when the inputted sentence is read aloud. Is predicted, and the breathing position is determined based on the predicted air discharge amount.

In the breath position determining method in the first sentence-speech conversion apparatus according to the present invention, the air discharge amount when the input sentence is read aloud is predicted, and the breath position is determined based on the predicted air discharge amount. Therefore, it is easy to determine the position where the actual breathing is performed when reading the input sentence as the breathing position, and by inserting a breathing sound or silence pause at the determined breathing position, a more natural voice can be obtained. You will get it.

A second breath-to-speech position determining method in the second sentence-speech conversion apparatus according to the present invention is a breath-to-speech position determining method in a sentence-to-speech conversion apparatus for converting an input sentence into a voice, based on the inputted sentence. A feature is that a voice signal corresponding to an input sentence is generated, and a breathing position is determined based on a cumulative value of voice power of the generated voice signal.

Specifically, for example, a breathing position candidate is determined based on the cumulative value of the voice power of a voice signal, and the break position in the grammar / semantic analysis that first appears after the determined breathing position candidate is breathed. Determine the position.

In the second breath-speech position determining method in the second sentence-speech conversion apparatus according to the present invention, since the breath-breath position is determined based on the cumulative value of the voice power of the voice signal corresponding to the input sentence, the input sentence is read aloud. The position where the actual breathing is performed at this time is easily determined as the breathing position, and a more natural voice can be obtained by placing a breathing sound or a silent pause at the determined breathing position.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

[1] Description of First Embodiment FIG. 1 shows a sentence-speech conversion apparatus according to a first embodiment of the present invention.

The sentence input to the sentence-speech converter is subjected to grammatical analysis, semantic analysis, etc. by the language processing unit 1,
It is sent to the voice generation unit 2. The voice generation unit 2 generates a voice signal according to the input sentence. The voice generation unit 2 outputs a voice signal having strength and weakness according to the content of the input sentence. As a method of adding strength to a voice, the language processing unit 1 separates a strongly uttered part and a weakly uttered part by semantic analysis based on an input sentence, and a strongly uttered part and weakly uttered part in advance in an input sentence. There is a method to put a code indicating the part to be done.

The voice signal generated by the voice generation unit 2 is sent to the breathing sound and silence insertion unit 3 and the breathing position candidate detection unit 5. The breath-and-sound and silence insertion unit 3 adds the sound signal generated by the sound generation unit 2 to the sound signal based on the control command from the breath-and-sound and silence insertion control unit 6.
Insert breath or silence audio signal. The voice signal obtained by the breath breath sound and silence insertion unit 3 is output as voice through the voice output unit 4.

The breathing position candidate detection unit 5 includes a voice generation unit 2
The power value of the sound is accumulated from the sound signal generated by, and when the accumulated value reaches a predetermined value, the breathing position information is sent to the breathing sound and silence insertion control unit 6. When the cumulative value reaches the predetermined value, the breathing position candidate detection unit 5 clears the cumulative value up to that point and newly starts calculation of the cumulative value.

When v (t) is a voice waveform, time points t1 to
Cumulative value P (t1, t) of the power value of the voice until time t3
2) is expressed by the following mathematical formula 1.

[0019]

[Equation 1]

The breathing sound and silence insertion control unit 6 receives the breathing position information from the breathing position candidate detecting unit 5 and also receives the punctuation mark position information and the meaning boundary position information extracted from the input sentence by the language processing unit 1. Sent.
Hereinafter, the punctuation mark position information and the meaning boundary position information are collectively referred to as delimiter position information in grammar / semantic analysis.

The breathless sound and silence insertion control unit 6 determines, based on the break position information (punctuation mark position information and semantic boundary position information) in grammatical / semantic analysis, and the breathing position information.
The following control is performed.

(1) The breathing sound insertion position is determined based on the breathing position represented by the breathing position information and the grammatical / semantic analysis dividing position represented by the grammatical / semantic analysis delimiter position information. , And controls the breathing sound and the silent insertion portion 3 to insert the breathing sound at the determined breathing sound insertion position.

That is, the breathing sound insertion position is determined to be a grammatical / semantic parsing position that appears first after the breathing position, out of the punctuation mark position and the meaning boundary position. The breathing position refers to the position of the audio signal corresponding to the time when the breathing position candidate detection unit 5 outputs the breathing position information.

(2) The silent insertion position is determined based on the breathing position represented by the breathing position information and the grammatical / semantic analysis delimiting position represented by the grammatical / semantic analysis delimiter position information, The breathing sound and silence insertion section 3 is controlled so as to insert the silence into the determined silence insertion position.

The silence insertion position is basically determined as a separation position in grammar / semantic analysis. However, if the breath insertion sound insertion position is present near the silence insertion position determined in principle, the determined silence insertion position is canceled. Also, in the vicinity of the silence insertion position that is determined in principle,
If there are other grammatical / semantic parsing positions, one of them is selected and determined as the silent insertion position.

By the way, when reading sentences, there are breath poses and breathless poses. Breath poses are relatively long and are important as semantic breaks, while breathless poses are relatively short and not necessarily important.

The timing of breathing at the time of reading a sentence is related to the lung residual capacity (residual respiratory capacity). That is, it is related to the amount of air discharged. In the above-described embodiment, the air discharge amount is calculated based on the power cumulative value of the sound generated by the sound generating unit 2, and when the power cumulative value reaches the predetermined value, the residual air volume of the lung is equal to or less than the predetermined value. It is determined that, and the first delimiter position in grammar and semantic analysis after that,
It has been determined to be a breathing position. Therefore, the position where the breath is breathed when the input sentence is read aloud is easily determined as the breathing position. In addition, at the detected breathing position, since the breathing sound is inserted instead of silence,
The input sentence can be converted into a more natural voice.

It should be noted that a table showing the relationship between the air ejection amount and the utterance amount is prepared for each phoneme, and the loudness (utterance amount) of each phoneme constituting the input sentence.
The air discharge amount may be calculated from the above, and the breathing position information may be generated based on the calculated cumulative value of the air discharge amount.

[2] Description of Second Embodiment

FIG. 2 shows a sentence-speech conversion apparatus according to the second embodiment of the present invention. 2, the same parts as those in FIG. 1 are designated by the same reference numerals and the description thereof will be omitted.

In this example, the breathing sound and silence insertion control unit 6 is provided with a timer 7 for measuring a preset time, and the method of determining the breathing sound insertion position is different from that of the sentence-speech conversion apparatus of FIG. There is.

The breathing position information output from the breathing position candidate detection unit 5 is referred to as first breathing position information. The breathing position candidate detecting unit 5 integrates the power value of the voice from the voice signal generated by the voice generating unit 2, and when the integrated value reaches a predetermined value, the first breathing position information is breathed. Send to the sound and silence insertion control unit 6.
In addition, the breathing position candidate detection unit 5 outputs the voice signal immediately after the breathing sound insertion position determined by the breathing sound and silence insertion control unit 6 when the integrated value reaches a predetermined value from the voice generation unit 2. In some cases, the integration value up to that point is cleared, and calculation of the integration value is newly started.

The timer 7 is reset when the voice signal immediately after the breathing sound insertion position determined by the breathing sound and silence insertion control unit 6 is output from the voice generating unit 2, and is reset and then reset. The second breathing position information is output when a predetermined time is counted until the time.

The breath insertion sound insertion position is, in principle, similar to the first embodiment, a delimiter position in grammatical / semantic analysis which first appears after the breathing position represented by the first breathing position information. The first breath sound insertion position) is determined.

However, when the second breathing position information is output before the first breathing position information is output,
Separation position in the grammatical / semantic analysis that first appears after the breathing position represented by the second breathing position information (second
The breathing sound insertion position) is determined as the breathing sound insertion position.

[0036]

According to the present invention, the input sentence can be converted into a more natural voice.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a sentence-speech conversion apparatus according to a first embodiment of this invention.

FIG. 2 is a block diagram showing a configuration of a sentence-speech conversion apparatus according to a second embodiment of this invention.

[Explanation of symbols]

 1 Language Processing Section 2 Voice Generation Section 3 Breathing Sound and Silent Insertion Section 4 Voice Output Section 5 Breathing Position Candidate Detection Section 6 Breathing Sound and Silent Insertion Control Section 7 Timer

Claims (6)

[Claims] 1. A sentence-speech conversion method for converting an input sentence into voice, wherein a breathing position is determined based on the input sentence, and a breathing sound is inserted at the determined breathing position. Sentence conversion method. 2. The sentence-speech conversion method according to claim 1, wherein the breathing position is determined based on a cumulative value of the voice power of the voice signal generated from the input sentence. 3. The breathing position is determined as the break position in the grammatical / semantic analysis first appearing after the breathing position candidate determined based on the cumulative value of the voice power of the voice signal generated from the input sentence. Claim 1 characterized by the above.
Sentence voice conversion method described in. 4. A breath position determining method in a sentence-speech conversion device for converting an input sentence into voice, comprising predicting an air discharge amount when reading the input sentence,
A breath-breathing position determining method in a sentence-speech conversion device, which determines a breath-breathing position based on a predicted air discharge amount. 5. A breath position determining method in a sentence-speech conversion device for converting an input sentence into a voice, wherein a voice signal corresponding to the input sentence is generated based on the input sentence, and the generated voice is generated. A breath-breathing position determining method in a sentence-speech conversion apparatus, which determines a breath-breathing position based on a cumulative value of voice power of a signal. 6. A breathing position candidate is determined based on a cumulative value of voice power of a voice signal, and a break position in grammatical / semantic analysis first appearing after the determined breathing position candidate is determined as a breathing position. The breathing position determining method in the sentence-speech conversion apparatus according to claim 5.