JP2821276B2 - 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 synthesizing apparatus for synthesizing speech, and more particularly to a speech synthesizing apparatus capable of increasing the utterance speed by an analysis and synthesis type apparatus.

[0002]

2. Description of the Related Art FIG. 2 shows the configuration of a conventional analysis-synthesis type speech synthesizer capable of increasing the utterance speed. In the speech synthesizer 11, a host CPU (central processing unit) 12 instructs an utterance speed and selects data of speech to be synthesized. Host CPU1
Reference numeral 2 denotes an instruction for inputting a speech unit to be synthesized into the speech data file 13. The voice data file 13 stores voice unit data, reads out voice unit data instructed, and outputs the read voice unit data to the voice data memory 14. Voice data memory 1
Reference numeral 4 denotes a storage medium for expanding speech unit data so as to be immediately accessible when requested by the host CPU 12.

The vocal tract characteristic parameters in the voice segment data stored in the voice data memory 14 are temporarily stored in a vocal tract characteristic parameter buffer 15, and the sound source parameters are temporarily stored in a sound source parameter buffer 16.
The buffer controller 17 controls these buffers 15,
This is for controlling the input / output of 16 data. The vocal tract characteristic parameter and the sound source parameter are supplied to the voice synthesis filter 18 under the control of the buffer controller 17, and the synthesized voice is output from the voice output terminal 19.

In such a conventional speech synthesizer 11,
In increasing the utterance speed, two techniques have been employed. In the first of these, the speech synthesis filter 18
The transfer speed of the vocal tract characteristic parameter and the sound source characteristic parameter to be transmitted to the TD is increased, and the period of D / A (digital / analog) conversion is increased. Further, in the second method, the speech characteristic parameters and the sound source parameters to be transferred to the speech synthesis filter 18 are thinned out in frame units, thereby increasing the utterance speed.

[0005]

In the first method, the utterance speed is simply increased by simply increasing the speed so as to perform fast forward reproduction of a music tape. For this reason, this method not only increases the utterance speed,
There is a problem that the pitch of the synthesized voice also increases, and the pitch of the voice increases as a whole.

[0006] The second method can solve the problem that the pitch of the synthesized speech becomes high. However, in the second method, in order to uniformly take out frames to be decimated, a frame such as a consonant section in which the utterance length does not change so much, or a frame such as a plosive which has importance in transient changes, is used. There was a possibility of thinning out even the frame. For this reason, if the utterance speed is increased,
There has been a problem that a loss of phoneme information occurs and the sound quality of synthesized speech is deteriorated.

An object of the present invention is to provide a speech synthesizer capable of realizing a high-speed synthesized speech without changing the pitch of the synthesized speech and without changing the sound quality of the synthesized speech. is there.

[0008]

According to the first aspect of the present invention,
Corresponds to the speech to be synthesized from the speech unit data (a)
An audio data memory for storing what
The segment data stored in the voice data memory is
Separation into data and cepstrum parameters
And (c) separated by this parameter separating means.
Sound source parameter buffer for storing sound source parameters
And (d) Kep separated by the parameter separation means
Cepstrum parameters that store strum parameters
Buffer and (e) cepstrum parameter buffer
Investigate the cepstrum parameters of
Higher order cepstrum of audio data in investigated frame
Whether there is a peak indicating the presence of pitch in the coefficient
Therefore, the pitch of the audio data in the investigated frame
Pitch determination means for determining absence, and (f) high utterance speed
Means for instructing to increase the utterance speed, and (g)
The voice speed increase instruction means instructs to increase the utterance speed.
Pitch determining means for activating the determination of the pitch determining means
Moving means and (h) the pitch determining means starting means
When the pitch determination means is activated, the pitch determination means
The sound based on the pitch period as the period between the determined pitches
Source parameter pitch is detected and the pitch pulse position
Pitch pulse position marking means for marking
(I) High speed indicated by the utterance speed increase instruction means
Of pitch pulse position marking means according to the speed of
Sound source parameters in data units between pitch pulses
High-speed sound source parameter decimation means for decimation;
(V) High utterance speed by means of high utterance speed instruction means
Decimated from sound source parameter buffer when activation is instructed
Sound source parameters and cepstrum parameter
Audio signal using cepstrum parameters from web
And a speech synthesis filter for synthesizing
Let

[0009] Then, the rate of thinning out of data in the pitch section is adjusted according to the speed of the synthesized voice to output a high-speed synthesized voice, thereby achieving the above object.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to embodiments.

FIG. 1 shows the configuration of a speech synthesizer according to an embodiment of the present invention. This speech synthesizer 2
Reference numeral 1 includes a voice data file 13 and a voice data memory 14 similarly to the conventional voice synthesizer 11 described with reference to FIG. The sound source parameters in the speech unit data are temporarily stored in the sound source parameter buffer 16, and the cepstrum parameters are stored in the cepstrum parameter buffer 22.
Is stored temporarily.

A host CPU 23 for instructing the utterance speed and selecting speech data to be synthesized is connected to the speech data file 13, the speech data memory 14, and the buffer controller 24. The buffer controller 24 is connected to a pitch determination unit 25, a pitch detection unit 26, and a speech synthesis filter 27 in addition to the sound source parameter buffer 16 and the cepstrum parameter buffer 22.

Here, the pitch determination unit 25 determines the presence or absence of a pitch component using cepstrum parameters. Further, the pitch detection unit 26 detects the pitch in the pitch section from the sound source parameters based on the determination by the pitch determination unit 25. The host CPU 23 selects data to be used for synthesis, controls the generation speed, and the like. The buffer controller 24 is a controller that controls input and output of data of the cepstrum parameter buffer 22 and the sound source parameter buffer 16. The voice synthesis filter 29 synthesizes voice using cepstrum parameters and sound source parameters, and outputs voice from the voice output terminal 19.

The operation of the speech synthesizer 21 having such a configuration will be described below. First, it is assumed that the host CPU 23 instructs utterance of voice at a normal speed. In this case, speech unit data for the speech to be synthesized is read from the speech data memory 14. The data set compiled for each frame is separated into cepstrum parameters and sound source parameters. The cepstrum parameters are stored in the cepstrum parameter buffer 22 and the sound source parameters are stored in the sound source parameter buffer 16. Then, voice synthesis is performed using these parameters, and voice is output from the voice output terminal 19. That is, in this case, the pitch determination unit 25 and the pitch detection unit 26 are not activated, the data in the frame is processed without being thinned, and the speech is synthesized.

When the host CPU 23 instructs to increase the utterance speed, the contents of the instruction are input to the buffer controller 24, and the pitch determination unit 25 is activated. The pitch determination unit 25 checks the parameters in the cepstrum parameter buffer 22 on a frame-by-frame basis, and determines whether there is a pitch in the sound data in the checked frame.

The determination of the presence or absence of the pitch is realized by checking whether or not a peak representing the presence of the pitch exists in the higher order coefficient of the cepstrum. If a pitch exists in the frame, the pitch detection unit 26 is activated. Then, pitch detection is performed for the sound source parameters of the same frame using the pitch period obtained by the pitch determination unit 25 as a clue, and the position of the pitch pulse is marked. If no pitch exists in the frame, the pitch determination unit 25 does not activate the pitch detection unit 26. Therefore, in this case, the sound source parameters as data in the frame are not thinned out.

In this way, the processing for speeding up the synthesized speech is performed in frame units. The buffer controller 24 determines the pitch marked by the pitch detection unit 26 based on the utterance speed specified by the host CPU 23. The higher the speed, the more the sound source parameters are thinned out and transferred from the sound source parameter buffer 16 to the speech synthesis filter 27 in units of data between pulses. Voice synthesis filter 27
Synthesizes an audio signal using the data of the frame unit transferred from the cepstrum parameter buffer 22 and the sound source parameter buffer 16 and outputs it from the audio output terminal 19.

[0018]

As described above , according to the first aspect of the present invention, the cepstrum separated by the parameter separating means.
The frame investigated by the pitch judgment means using the parameters
The pitch of the voice data in the system
Control is achieved by thinning out sound source parameters in pitch units
doing. Therefore, the configuration of the speech synthesizer is simple
Not only is the processing when the utterance speed is increased
It is simple and does not burden the processing system for this.
There is no effect.

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁶ , DB name) G10L 9/00 G10L 9/16

Claims (1)

(57) [Claims] 1. A method for synthesizing speech from speech segment data.
An audio data memory for storing the corresponding data, and the segment data stored in the audio data memory are stored in a sound source pattern.
Parameters separated into parameters and cepstrum parameters
Data separating means and sound source parameters separated by the parameter separating means.
And a cepstrum separated by the parameter separating means.
Cepstrum parameter buffer for storing system parameters
And the cepstrum in the cepstrum parameter buffer.
The parameters are investigated in frame units, and the investigated frame
Pitch to the higher order coefficient of the cepstrum of the audio data in the
Investigate whether there is a peak indicating the presence of
To determine the presence or absence of the pitch of the audio data in the
Switch determination means and utterance speed increase instructing means for instructing to increase the utterance speed
And the utterance speed increase instructing means instructs to increase the utterance speed.
Pitch that activates the judgment of the pitch judgment means when
Determining means activating means, and the pitch determining means
The pitch determined by this pitch determination means when the stage is activated
The sound source parameters based on the pitch period as the period between
Data pitch and mark the position of the pitch pulse.
Pitching pulse position marking means to be executed, and a speed increase instructed by the utterance speed increase instructing means.
Mark of pitch pulse position marking means according to the speed of
Sound source parameters in data units between clicked pitch pulses.
High-speed sound source parameter thinning-out means for thinning out data,
From the sound source parameter buffer when
Sound source parameters and the cepstrum parameters
Sound using cepstrum parameters from
A speech synthesis device comprising: a speech synthesis filter that synthesizes a voice signal .