JP3113101B2 - Speech synthesizer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech synthesizer for generating a synthesized speech from a character code string or a prosody information and a phoneme sequence.

[0002]

2. Description of the Related Art Recently, various speech synthesizers have been developed which analyze a sentence containing kanji and kana characters and synthesize and output speech information indicated by the sentence by a rule synthesis method. This type of speech synthesizer has begun to be widely used as a telephone introduction service in a banking business, a newspaper review system, a text-to-speech device, and the like.

A speech synthesizer employing this type of rule synthesizing method basically converts speech uttered by a human into a predetermined unit,
For example, for each of CV (consonant, vowel), CVC (consonant, vowel, consonant), VCV (vowel, consonant, vowel), VC (vowel, consonant), LSP (line spectrum pair) analysis and cepstrum analysis are used. The phoneme information obtained by the analysis is registered in a speech unit file, and synthesis parameters (phoneme parameters and prosodic parameters) are generated with reference to the speech unit file, and a sound source is generated based on these synthesis parameters. And a synthetic filtering process to generate synthesized speech.

[0004] Further, such a speech synthesizer can freely change the utterance speed without changing the pitch height. Conventionally, as a method for changing the utterance speed, a method of increasing or decreasing the length (mora length) of a single syllable (CV or C) when generating a synthesis parameter, or a method of increasing or decreasing a frame period when synthesizing is generated. Are known.

The utterance speed control method based on the mora length (variable) controls the sound by adopting a feature that when a utterance speed is changed by a human, the consonant is almost constant and only the vowel portion expands and contracts. For this reason, in the utterance speed control method based on the mora length, when attempting to synthesize speech at such a speed that even humans cannot utter, the vowel part becomes too short or the vowel part disappears, which makes it very difficult to hear. I was As a means for solving this, a method of uniformly thinning out synthesis parameters according to the mora length is known to increase the utterance speed.However, control for thinning out frames is difficult, and clarity is reduced due to thinning. And the connection of the sound becomes poor.

On the other hand, the utterance speed control method based on (variable) frame period can expand and contract vowels and consonants uniformly, and when the utterance speed is increased, the vowel part cannot be heard. Since there is no need to skip frames, the connection between sounds does not deteriorate. However, in the utterance speed control method based on the frame period, when the utterance speed is reduced, the consonant becomes longer as well as the vowel, so that there is a problem that the sound sounds unnatural.

[0007]

As described above, in the above-mentioned conventional speech synthesis technology, if the speech speed control method based on the variable mora length and the speech speed control method based on the variable frame period are individually used, the speech speed depends on the speech speed. There were inconveniences such as sounds that were difficult to hear and unnatural sounds.

The present invention has been made in view of such circumstances, and it is an object of the present invention to be able to generate a synthesized sound that is easy to hear even when the utterance speed is extremely high, and to provide a natural synthesized sound even when the utterance speed is low. An object of the present invention is to provide a speech synthesizer capable of generating sound.

[0009]

According to the present invention, there is provided a speech synthesizer comprising: an input unit for inputting speech speed information of a synthesized speech;
The speech rate information input to the input unit is determined in advance.
Indicates that the text should be read faster than
Means for determining whether the
In determining means, that speak faster than said input utterance speed loach report is stipulated Me premixed value is determined to be the table
When the first means of shortening the frame period during speech synthesis, by the determining means, and the table that is read aloud slower than the speech speed information the input <br/> predetermined value
And a second means for generating a composite parameter by extending the mora length when it is determined that

[0010]

In the above arrangement, the input speech speed information
Determine if is possible to increase the speech rate is designated by the broadcast
In the case of rejection, it is possible to shorten consonants and vowels uniformly by shortening the frame period, and conversely, if it is determined that the speech speed is specified to be low, By increasing the length, it is possible to extend only the vowels, so that changing the speech rate does not cause intelligibility or unnaturalness, and high-quality synthesized speech can be generated. Becomes

[0011]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of the speech synthesizer according to the embodiment. The speech synthesizer shown in FIG.
It has an input unit 1 for inputting a character code string mixed with kanji or kana to be subjected to speech synthesis and speech speed information x of synthesized speech.

The speech synthesizing apparatus shown in FIG. 1 also has a word dictionary 2 in which accent types, readings, parts of speech, and the like for words and phrases to be subjected to speech synthesis are registered in advance, and an input unit 1 inputs the words. And a language processing unit 3 that analyzes the character code string using the word dictionary 2 and generates a corresponding phoneme sequence and prosody information.

The speech synthesizer shown in FIG. 1 also includes a speech unit file 4 in which a phoneme parameter group obtained by analyzing an input speech in advance for each arbitrary speech unit is stored. The phoneme parameters corresponding to the phoneme sequence generated by the above are extracted from the speech unit file 4,
A synthesis parameter generation unit 5 that determines mora lengths of various phonemes based on a mora length determined by (a mora length setting unit 7 in (a mora length setting unit 7 in) an after-mentioned speech speed control unit 10 and generates phoneme parameters). The synthesis parameter generation unit 5 also generates a prosody parameter according to the prosody information generated by the language processing unit 3.

The speech synthesizer shown in FIG.
The speech rate control unit 10 controls the speech rate by setting the mora length and the frame period based on the speech rate information x input by the control unit 1 and the synthesis parameters (phonological parameters and A speech synthesis unit 11 for generating a sound source and performing digital filtering in accordance with a prosody parameter) and a frame cycle determined by the speech rate control unit 10 (a frame cycle setting unit 9 described below) to generate a synthesized speech; Speaker 12.

The speech rate control unit 10 uses the mora length table 6 in which the correspondence between the speech rate information (the value thereof) and the mora length is registered, the speech rate information x from the input unit 1 and the mora length table 6, and Using a mora length setting unit 7 for determining the length, a frame period table 8 in which the correspondence between (the value of) the speech speed information and the frame period is registered, and the speech speed information x and the frame period table 8 from the input unit 1. It has a frame period setting unit 9 for determining a frame period. In FIG. 1, a D / A converter for converting a synthesized voice into an analog signal in the voice synthesis unit 11 and the like are omitted.

Next, the operation of the speech synthesizer shown in FIG. 1 centered on the speech rate control section 10 will be described with reference to the flowchart of FIG. Here, the speech speed information is defined as “(length of a single syllable of speech) / 10”.

First, when the utterance speed information x is input from the input unit 1, the utterance speed information x is input to the utterance speed control unit 10.
And the information x (the value of)
And (a value of) predetermined reference speed information, for example, “12” (Steps S1 and S2).

As a result of this comparison, if x is less than 12, the mora length setting unit 7 in the speech rate control unit 10 determines the mora length to be 12 frames, and
(Step S3). The speech rate control unit 10
The frame period setting unit 9 determines the frame period corresponding to the information x with reference to the frame period table 8 based on the input speech speed information x, and sets the frame period in the speech synthesis unit 11 (step S4).

On the other hand, when the speech rate information x from the input unit 1 is 1
If it is 2 or more, the frame cycle setting unit 9 determines the frame cycle of a normal value (here, the frame cycle at the time of analysis, for example, 10 ms) and sets it in the speech synthesis unit 11 (step S5). Further, the mora length setting unit 7 refers to the mora length table 6 based on the input speech speed information x, determines a mora length corresponding to the information x, and sets the mora length in the synthesis parameter generation unit 5 (step S6). The operation of the above speech synthesizer will be described more specifically.

First, it is assumed that utterance speed information x having a value of "12" is input by the input unit 1. Here, the sampling frequency at the time of analysis (for obtaining phoneme information to be registered in the speech unit file 4 from a voice generated by a human) is "8
kHz (125 μs) ”and the frame cycle is“ 10 ms ”
And Also, FIG.
The content shown in FIG. 3A is used, and the content shown in FIG.

The speech speed information x = 12 input by the input unit 1 is taken into the speech speed control unit 10 (step S1), and the processing is performed according to the flow shown in the flowchart of FIG. First, the input speech speed information x = 12
Is the reference utterance rate information "12" in step S2.
Is compared to In this case, since x = 12, step S
Steps S5 and S6 are performed. In step S <b> 5, a frame cycle of “10 ms” is determined by the frame cycle setting unit 9 and set in the speech synthesis unit 11. In step S6, the mora length table 6 shown in FIG. 3A is referred to by the mora length setting unit 7, and the speech speed information x =
The mora length “12” corresponding to “12” is set in the synthesis parameter generation unit 5. That is, in this example, since the frame period is “10 ms” and the mora length is “12 frames”, the length per mora is “120 ms” (10 ms).
ms × 12).

Here, when the input unit 1 inputs a character code string mixed with kanji or kana to be subjected to speech synthesis,
The language processing unit 3 collates the input character code string with the word dictionary 2 and obtains accent type, reading, and part of speech information of a word or phrase which is a target of speech synthesis indicated by the input character code string. Then, it determines the accent type and boundary according to the part-of-speech information, and converts the sentence into a form of reading a sentence mixed with kanji and kana to generate a phoneme sequence and prosodic information.

The phoneme sequence and the prosody information generated by the language processing unit 3 are provided to a synthesis parameter generation unit 5. The synthesis parameter generation unit 5 extracts a phoneme parameter corresponding to the phoneme sequence given from the language processing unit 3 from the speech unit file 4, and sets the mora length “set by the mora length setting unit 7 in the speech speed control unit 10. Based on “12”, that is, 12 frames, mora lengths of various phonemes are determined, and phoneme parameters are generated. At the same time, the synthesis parameter generator 5
Generates a prosody parameter according to the prosody information given from the language processing unit 3. That is, the synthesis parameter generation unit 5
Generates a synthesis parameter including a phoneme parameter and a prosody parameter.

The synthesis parameters generated by the synthesis parameter generator 5 are given to the speech synthesizer 11. The voice synthesizing unit 11 controls the synthesis parameters and the speech speed control unit 1
A sound source is generated and digital filtering is performed according to the frame period set by the frame period setting unit 9 within 0 to generate a synthesized voice.

Here, the phoneme parameters (speech units) extracted from the speech unit file 4 by the synthesis parameter generation unit 5 are converted into consonants (C) as shown in FIG.
It is assumed that the ratio of the length of the vowel (V) is “C: V = 6: 6” and the mora length is 12 frames. On the other hand, a frame cycle “10 ms” is set in the speech synthesis section 11 by the frame cycle setting section 9. In this case, the voice synthesizing unit 11 has a frame cycle of “10 ms” and a sampling cycle of “125 μs”, so that 80 (10 ms / 125 μs) for one frame of synthesis parameters.
Will be obtained.

Next, the case where the speech rate information x having the value "6" is inputted by the input unit 1 will be described. In this case, steps S1, S2, S3, and S4 in FIG. 2 are processed to obtain the mora length “12” and the frame corresponding to the speech speed information x = 6 from the frame period table 8 shown in FIG. A period “5 ms” is obtained. Therefore, the length per mora is “60 ms” (5 ms × 1
2) becomes 40 for the synthesis parameter of one frame.
(5 ms / 125 μs) of audio waveform data will be generated. That is, as shown in FIG.
Is “6: 6” and 60 m from 120 ms in FIG.
s.

Next, the case where the speech rate information x having the value "20" is inputted by the input unit 1 will be described. In this case, steps S1, S2, S5, and S6 in FIG. 2 are processed to obtain a frame period “10 ms”,
From the mora length table 6 shown in FIG.
The mora length “20” corresponding to 20 is obtained. Therefore, the length per mora is “200 ms” (10 ms
× 20), and 80 (10 ms / 125 μs) audio waveform data are generated for one frame of synthesis parameters. That is, as shown in FIG.
Only C: V becomes “6:14”. In addition,
Decompression of V is performed in the last frame of the speech unit, or V
May be performed by repeatedly using the frame in the middle part of the above.

As described above, according to the apparatus of the present embodiment having the above-described processing function, when the utterance speed is higher than the predetermined speed "12", the frame period is shortened. To synthesize a voice, and conversely, a predetermined speed “1”
If the speed is slower than "2", the mora length is lengthened and speech synthesis is performed, so that a synthesized speech that is easy to hear can be generated even when the speech speed is changed.

The present invention is not limited to the above embodiment. That is, in the above-described embodiment, the switching control between the two methods is performed based on the speech speed information “12” and the mora length corresponding to the speech speed information “12” is “12 frames”. It is not limited.
Further, when the utterance speed information is less than “12”, the mora length is fixed at “12”, but may be set to be gradually shortened. Similarly, when the frame period is “12” or more, the frame period is fixed to “10 ms”, but may be set to be gradually longer. In short, the present invention can be variously modified and implemented without departing from the gist thereof.

[0030]

As described above, according to the present invention, according to the value of the speech speed information, one of the speech speed control system based on the variable frame period and the speech speed control system based on the variable mora length is selected or used in combination. Since the disadvantages of each of the above methods can be compensated for by the other method, even when the speech speed is greatly changed, it is possible to produce a synthetic voice that is natural and easy to hear.

[Brief description of the drawings]

FIG. 1 is a block diagram of a speech synthesizer according to an embodiment of the present invention.

FIG. 2 is a flowchart for explaining a processing flow of an utterance speed control unit 10 in the embodiment.

FIG. 3 is a view showing an example of contents of a mora length table 6 and a frame cycle table 8 in the embodiment.

FIG. 4 is a diagram showing a change in the length of a CV syllable due to a difference in the value of speech speed information.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Input part, 2 ... Word dictionary, 3 ... Language processing part, 4 ... Speech unit file, 5 ... Synthetic parameter generation part, 6 ... Mora length table, 7 ... Mora length setting part, 8 ... Frame period table, 9 ... Frame period setting unit, 10 ... Speaking speed control unit, 11 ... Voice synthesis unit.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-62-223697 (JP, A) JP-A-4-170600 (JP, A) JP-A-64-79799 (JP, A) JP-A-3- 203800 (JP, A) JP-A-3-206496 (JP, A) JP-A-62-233831 (JP, A) JP-A-61-122700 (JP, A) JP-A-59-200340 (JP, A) JP-A-58-16295 (JP, A) JP-B-56-2354 (JP, B2) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 11/00-13/08 G10L 19/00 -21/06

Claims (1)

(57) [Claims] 1. A generates a prosodic parameters according prosody information to generate a corresponding phoneme parameters according phoneme sequence, in the speech synthesizing apparatus for synthesizing speech in accordance with synthesis parameters consisting phoneme parameters and prosodic parameters, the speech speed of the synthesized speech Input where information is entered
Unit and speech rate information input to the input unit are determined in advance.
Indicates that the text should be read faster than
Determining means for determining whether the means that increase, in this determination means, input to the input unit utterance speed soil report
Have tables to but that read aloud faster than value determined Me pre
When it is determined that that the first means for shortening the frame period during speech synthesis, in the determination unit, input to the input unit utterance speed soil report
Not table a but it is read aloud slower than the value that is determined Me premixed
And a second means for generating a synthesis parameter by extending the mora length when it is determined that the speech is synthesized.