JPH096377A - Voice synthesizer and its control method - Google Patents

Abstract

PURPOSE: To make it possible to edit the phonetic data of the result of voice language processing and to realize various vocalization for an input sentence of the same character string. CONSTITUTION: A voice language processing part 103 voice-language processes for an inputted sentence from a sentence input part 101, and generates the phonetic data for voice synthesis to store them in a phonetic data hold part 104. A phonetic data hold part 105 displays the phonetic data, and a phonetic data correction part 106 performs the required correction of the displayed phonetic data. The corrected phonetic data are held at a phonetic data correction value hold part, and the voice is generated using the phonetic data after correction. In such a manner, since synthesized voice is outputted by a parameter based on the phonetic data after correction, a synthesized voice rich in variety can be generated by correcting the phonetic data even for the same character string.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice synthesizing apparatus and a control method thereof, and more particularly to a voice synthesizing apparatus suitable for performing voice synthesizing processing based on accent phrase unit information and a control method thereof.

[0002]

2. Description of the Related Art Generally, in a speech synthesizer, an input sentence is analyzed and divided into accent phrase units, and acoustic information in which information such as accent and pause is added is represented by text or a frame (hereinafter referred to as a phonetic symbol). (Referred to as data) is input to the sound processing unit.

This phonetic data is based on the result of sentence analysis processing such as morphological analysis and phrase relation analysis, and includes pause information (a range that occurs at one breath (referred to as a pause phrase), silence periods placed between pause phrases). Length), reading (corresponding to changes in reading), accent generation (combination of accents), and other speech processing to generate synthetic speech based on information obtained from language analysis. Is the result of

FIG. 4 shows an example of phonetic data for an input sentence "Today is good weather." In the example of FIG.
Except for the 20ms pause after "Today"
The standard value "5" is given for the speaking speed, pitch, and volume.

The prosody control section analyzes (interprets) this phonetic data, extracts information such as accent information, pause information, and pitch frequency necessary for generating prosody information therefrom, and generates a pitch pattern. . Further, the voice generation unit generates parameters of the synthetic voice from the pitch pattern, and the DA converter generates the synthetic voice.

As described above, the phonetic data includes acoustic information such as phoneme information, accent information, and control information.

[0007]

However, the above-mentioned conventional speech synthesizer has a drawback that it can generate only one kind of synthesized speech for an input sentence having the same face.
In other words, when you want to emphasize a specific phrase (for example, "Today" or "Good" in Fig. 4) or make the whole sentence into suspicious tone (upward) in a sentence of the same letter. There is. In general, to emphasize a particular phrase, there are various correction factors such as increasing the volume, slowing the speaking speed, increasing the pitch, etc.There is no definitive method to realize with a computer. .

The present invention has been made in view of the above problems, and makes it possible to edit phonetic data as a result of speech language processing, and to perform prosody control and speech generation using the edited values. It is an object of the present invention to provide a speech synthesizer and a control method thereof that realize various utterances for input sentences of the same character.

Another object of the present invention is to retain a set of edited phonetic data values so that they can be reused.

[0010]

A speech synthesis apparatus of the present invention for achieving the above object has the following configuration. That is, a generation unit that performs voice language processing on a sentence that performs voice synthesis to generate phonetic data for voice synthesis, and a correction unit that performs desired correction on the phonetic data generated by the generation unit. , Generating means for generating the parameters of the synthetic voice based on the phonetic data modified by the modifying means.

Further, preferably, the correction means inputs a display means for displaying the phonetic data generated by the generating means, and a desired correction instruction for the phonetic data displayed by the display means. Input means for This is because the phonetic data can be displayed and the displayed phonetic data can be modified, which facilitates the modification work.

Further, preferably, the generating means generates phonetic data of a sentence for which voice synthesis is performed in units of pause phrases.
This is because the phonetic data can be modified for each pause phrase, and various modifications are possible.

Further, preferably, the phonetic data modified by the modifying means includes information for determining the pose length, vocalization speed, pitch, volume and accent type of each pause phrase. These pieces of information are effective in emphasizing a desired pause phrase and making a sentence into a question tone.

Further, preferably, the apparatus further comprises holding means for holding the set of phonetic sound data corrected by the correcting means, and the correcting means reads out the set of phonetic data held by the holding means and stores the set. A means for correcting the phonetic data generated by the generating means is included. This is because the correction values created in the past can be reused and the phonetic data can be easily corrected. For example, if a set of correction values of phonetic data for emphasis is held, then the phonetic data for emphasis can be corrected by using the held correction values.

Further, preferably, the holding means holds a set of phonetic data in a state in which a unique label is provided, and the correcting means designates a label to hold the phonetic sound held in the holding means. Means for acquiring a set of data and modifying the phonetic data generated by the generating means by the acquired set of phonetic data. By adding a label to a set of correction values, it becomes possible to easily specify a desired correction value even when a plurality of sets of correction values are held.

A method of controlling a speech synthesizer according to the present invention for achieving the above object is a method of controlling a speech synthesizer for performing speech synthesis based on an inputted sentence, wherein a sentence for speech synthesis is used. A generation step of performing speech language processing on the phonetic sound data to generate phonetic data for voice synthesis, a correction step of making desired corrections to the phonetic data generated in the generation step, and a correction step corrected by the correction step. And a generation step of generating a parameter of the synthetic voice based on the phonetic data.

[0017]

According to the above construction, in the voice synthesizing device for synthesizing the voice based on the inputted sentence, the phonetic data for the voice synthesizing is generated by performing the voice language processing on the sentence for the voice synthesizing. To be done. Then, the phonetic data is modified as desired, and the parameters of the synthetic voice are generated based on the modified phonetic data.

Since the synthesized voice is output according to the parameters generated as described above, it is possible to generate a variety of synthetic voices by modifying the phonetic data even for the same character. Become.

[0019]

Embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram showing the schematic arrangement of the speech synthesizer of this embodiment. In the figure, reference numeral 1 is a CPU, which performs various controls in the speech synthesizer. A ROM 2 stores a control program executed by the CPU 1. The ROM 2 also stores a control program for executing the control described in the flowchart described later. A RAM 3 provides a work area for the CPU 1 to execute various controls. A voice output unit 4 generates and outputs voice based on the synthesized voice signal.

An external memory 5 stores a correction pattern of phonetic data for voice synthesis processing. An input unit 6 inputs a sentence to be voice-synthesized and a correction value of phonetic data. 7 is a display unit, CR
T, a liquid crystal display, etc., and displays phonetic data and the like. Reference numeral 8 denotes a bus, which connects the above-mentioned components to each other and enables data exchange between the components.

FIG. 2 is a block diagram showing the functional arrangement of the speech synthesizer of this embodiment. In the figure, reference numeral 101 denotes a sentence input unit for inputting a sentence for voice synthesis from the input unit 6. Reference numeral 102 denotes the input sentence input from the sentence input unit 101
It is an input sentence holding unit (buffer) held in AM3.

Reference numeral 103 denotes a speech language processing unit which performs speech language processing on the input sentence and generates phonetic data such as speech notation, accent type, pause length, utterance speed, and volume. Reference numeral 104 denotes a phonetic sound data holding unit (buffer) that holds the phonetic sound data generated by the voice language processing unit 103 in the RAM 3.

Reference numeral 105 denotes a phonetic data display section for displaying the value of the phonetic data held by the phonetic data holding section 104 on the screen of the display section 7. Reference numeral 106 denotes a phonetic data correction unit for correcting and inputting the value of the phonetic data displayed on the phonetic data display unit 105 using the input unit 6. Reference numeral 107 denotes a phonetic data correction value holding unit (buffer) which holds the value corrected and input from the phonetic data correction unit 106 in the RAM 3.

Reference numeral 108 denotes a correction pattern label input unit for inputting a label for managing the set of values held in the phonetic data correction value holding unit 107 as a pattern using the input unit 6. A correction pattern label holding unit 109 holds the pattern label input from the correction pattern label input unit 108 in the RAM 3. Reference numeral 111 denotes a correction pattern holding execution unit that holds the value held in the phonetic data correction value holding unit 107 and the label held in the correction pattern label holding unit 109 in the correction pattern holding unit 110 of the external memory 5 in association with each other. is there. The correction pattern holding execution unit is activated by inputting an execution instruction from the input unit 6.

Reference numeral 112 is a use pattern label input section for inputting the label of the correction pattern to be used from the input section 6. Reference numeral 113 is a use pattern label holding unit (buffer) that holds the use pattern label input from the use pattern label input unit 112 in the RAM 3. Reference numeral 114 reads the correction pattern corresponding to the label of the use pattern held in the use pattern label holding unit 113 from the correction pattern holding unit 110, and the phonetic data correction value holding unit 107.
It is a usage pattern reading execution unit held in. The usage pattern reading execution unit 114 is activated by inputting an execution instruction from the input unit 6.

Reference numeral 115 is a prosody control unit for controlling prosody information such as pause length, pitch, and volume by referring to the values held in the phonetic data correction value holding unit 107. A prosody information storage unit (buffer) 116 stores the prosody information obtained by the prosody control unit 115 in the RAM 3. Reference numeral 117 is a voice generation unit that performs voice generation processing such as prosody, parameter connection, and waveform connection by referring to the value stored in the prosody information storage unit 116. Reference numeral 118 is a voice parameter holding unit (buffer) that holds the parameters of the synthetic voice generated by the voice generating unit 117 in the RAM 3. A DA converter 119 converts the voice parameter held in the voice parameter holding unit 118 into synthetic voice. Also, 120 is the input unit 6
By inputting an execution instruction from, the prosody control unit 115 controls the prosody control and the voice utterance unit 117 controls the start of execution of voice generation.

Next, the operation of this speech synthesizer will be described.
FIG. 3 is a flow chart showing the operation procedure of the speech synthesizer of this embodiment.

In FIG. 3, steps S201 to S20 are performed.
7 is a state of waiting for input. The state of waiting for input is repeated until one of sentence input, phonetic data correction, correction pattern label input, correction pattern holding execution, use pattern label input, use pattern reading execution, and utterance execution is input.

When a sentence for speech synthesis is input to the sentence input unit 101, the input sentence is held in the input sentence holding unit 102, and the process proceeds from step S201 to step S208. In step S208, the speech language processing unit 103 performs normal speech language processing on the input sentence held in the input sentence holding unit 102, and holds the phonetic data as a result in the phonetic data holding unit 104. Then, in step S209, the values of the phonetic data such as the pause, pitch, and power held in the phonetic data holding unit 104 are displayed in the phonetic data display unit 1.
Indicated by 05. Then, the process is step S20.
Move to 1 and shift to the input waiting state.

When the correction value of the phonetic sound data is input from the phonetic data correction unit 106, the process proceeds from step S202 to step S210. In step S210, when the value is held in the phonetic sound data holding unit 104, the process proceeds to step S211, and the correction value input from the phonetic data correction unit 106 is held in the phonetic data correction value holding unit 107. On the other hand, if the phonetic data is not held in step S210, the process proceeds to step S222, and, for example,
Output (display or voice output) the error message "The sentence has not been entered. Please enter it.",
It returns to step S201.

When a label is input from the correction pattern label input unit 108, steps S203 to S2 are performed.
Move to 12. In step S212, the label of the correction pattern input from the correction pattern label input unit 108 is held in the correction pattern label holding unit 109, and in Step S212.
Move to 201.

When a correction pattern holding execution instruction is input from the correction pattern holding execution unit 111, step S20 is executed.
Then, the process proceeds from step 4 to step S213. In step S213, it is checked whether the correction value and the label are held in the phonetic data correction value holding unit 107 and the correction pattern label holding unit 109, respectively. If the correction value and the label are held, the process proceeds to step S214, and the correction value held in the phonetic data correction value holding unit 107 is stored in the correction pattern holding unit 110 together with the label held in the correction pattern label holding unit 109. Hold. Then, the process returns to step S201. If at least one of the correction value and the label is not held in step S213, the process proceeds to step S223, and, for example, an error message "The correction value or label has not been entered. Please enter." Is output. , Go to step S201.

When the label of the pattern to be used is input from the use pattern label input unit 112, step S
The procedure moves from 205 to step S215. Step S215
Then, the label input from the usage pattern label input unit 112 is held in the usage pattern label holding unit 113, and the process proceeds to step S201.

When an instruction to read and execute a pattern to be used is input to the use pattern read execution unit 114,
The process moves from step S206 to step S216. In step S216, it is determined whether or not the label to be used is held in the use pattern label holding unit 113, and if held, the process proceeds to step S217. And step S
In 217, the correction pattern corresponding to the label held in the usage pattern label holding unit 113 is read from the correction pattern holding unit 110 and held in the phonetic data correction value holding unit 107. Then, it returns to step S201. If the label used in step S216 is not held, the process proceeds to step S224.
For example, output the error message "The label has not been entered. Please enter it."
Return to 201.

When the instruction to execute the utterance is input from the utterance executing unit 120, the process proceeds from step S207 to step S218. In step S218, it is determined whether or not the phonetic data holding unit 107 holds the phonetic data, and if the phonetic data is held, the process proceeds to step S219.

In step S219, the prosody control unit 115
In reference to the phonetic data held in the phonetic data holding unit 104 and the correction values held in the phonetic data correction value holding unit 107, prosody information such as pause, pitch, and power is controlled. The prosody information storage unit 116 stores the prosody information.
To hold. Subsequently, in step S220, the voice generation unit 117 performs voice generation processing such as prosody, parameter connection, and waveform connection with reference to the value held in the prosody information holding unit 116, and the resulting voice parameter is obtained. Is stored in the voice parameter storage unit 118. Finally, in step S221, the DA converter 119 converts the voice parameter held in the voice parameter holding unit 118 into a synthesized voice, outputs the synthesized voice, and then in step S201.
Return to On the other hand, if the data is not held in the phonetic data holding unit 104 in step S218, the process proceeds to step S225, for example, an error message "Sentence has not been entered. Please enter" is output, Then, the process proceeds to step S201.

According to the control procedure as described above, it is possible to make desired corrections to the phonetic data obtained by performing the speech language processing on the input sentence. Therefore, it becomes possible to change the utterance speed and volume for each pause phrase. An example of correcting phonetic data is shown below.

FIG. 4 is a diagram showing an example of phonetic data for an input sentence "Today is a fine weather." FIG.
FIG. 5 is a diagram showing an example of correction of phonetic data in the case of emphasizing a pause phrase “today” in the input sentence of FIG. 4.
As a typical correction method for emphasizing a pause phrase, increase the pause length of the pause phrase (corrected from 20 to 40 in the figure), increase the volume (corrected from 5 to 7 in the figure), and increase the vocalization speed. It can be slowed down (corrected from 5 to 3 in the figure). In the example of FIG. 5, the pause phrase “today is” is emphasized when uttering.

Further, FIG. 6 shows that for the input sentence of FIG.
It is a figure which shows the correction example of phonetic data in the case of emphasizing the pose phrase "good." As in the case of FIG. 5, “good”
In the phonetic data corresponding to the pause phrase, the pause length is increased, the pitch is increased, and the utterance speed is decreased.

Further, FIG. 7 is a diagram showing an example of correction of phonetic data when the input sentence of FIG. 3 is interrogated. In this case, with respect to the phonetic data of the last pause phrase (“weather”), the accent type is set to 6 and the pitch at the end of the sentence is raised to make the interrogative utterance.

As described above, the phonetic data resulting from the speech language processing is edited, and the edited values are used to perform prosody control and voice generation, whereby input sentences having the same character face are A variety of vocalizations can be realized. Further, the edited value is reused by designating and holding the edited value as a pattern and designating a desired modification pattern with the label to read the pattern.

Incidentally, in the above-mentioned embodiment, in order to simplify the explanation, it is explained that the phonetic data is always corrected.
The synthetic voice may be generated without modification. In that case, a method of directly designating and using the value held in the phonetic data holding unit 104, holding the value in the phonetic data holding unit 104 and holding the value in the phonetic data correction value holding unit 107 at the same time A method, or a method in which the prosody control unit 115 uses the value held in the phonetic sound data holding unit 104 when the phonetic data correction value holding unit 107 has no value, can be considered.

Further, in the above embodiment, even if the data held in the phonetic data correction value holding unit 107 is all the phonetic data after correction, the difference from the value held in the phonetic data holding unit 104. It is okay if only.

Further, in the above-mentioned embodiment, the description has been made such that the set of corrected values of the phonetic data is held in one set of buffers at the same time, but a set of a plurality of values can be held in the buffer and switched. You may do so.

Further, in the above embodiment, the input sentence holding unit 10
Although it has been described that the speech language processing is performed immediately after the sentence is held in 2, the speech language processing may be executed by another input instruction.

Further, in the above embodiment, the prosody information holding unit 1
Although it has been described that the voice generation process is performed as soon as the sentence is held in 16, the voice generation process may be executed by another input instruction.

Further, in the above-described embodiment, the description is made such that all the input is performed manually, but other outputs such as a natural language generation device or a natural language dialogue device may be used as the input as they are.

As described above, according to the present invention, the phonetic data which is the result of the phonetic language processing unit 103 performing the phonetic language processing on the input sentence is displayed on the phonetic data display unit 105. , The prosody control unit 1 using the values edited by the phonetic data modification unit 106.
Since prosody control by 15 and voice generation processing by the voice generation unit 117 are performed, there is an effect that various synthetic voices can be uttered by correcting the phonetic data.

Further, the set of values edited by the phonetic data correction unit 106 is labeled as a pattern and held in the correction pattern holding unit 110, and the label input from the use pattern label input unit 112 is used to By reading the set of edited phonetic data values, there is an effect that the previously edited phonetic data values can be reused.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Needless to say, the present invention can be applied to a case where the present invention is achieved by supplying a program to a system or an apparatus.

[0052]

As described above, according to the present invention,
It is possible to edit the phonetic data as a result of the speech language processing, and by performing the prosody control and the voice generation using the edited values, various utterances can be realized for the input sentence of the same character.

Further, according to the present invention, a set of edited phonetic data values can be held and reused, and the table data can be corrected more easily.

[0054]

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of a speech synthesizer of this embodiment.

FIG. 2 is a block diagram showing a functional configuration of a speech synthesizer of this embodiment.

FIG. 3 is a flowchart showing an operation procedure of the speech synthesizer of this embodiment.

FIG. 4 is a diagram showing an example of phonetic data for an input sentence “Today is a good weather”.

5 is a diagram showing an example of correction of phonetic data in the case of emphasizing a pause phrase "today" in the input sentence of FIG.

FIG. 6 is a diagram showing a correction example of phonetic data in the case of emphasizing a pause phrase of “good” with respect to the input sentence of FIG. 3;

FIG. 7 is a diagram showing an example of correction of phonetic data when the input sentence of FIG.

[Explanation of symbols]

 101 sentence input unit 102 input sentence holding unit 103 speech language processing unit 104 phonetic data holding unit 105 phonetic data display unit 106 phonetic data correction unit 107 phonetic data correction value holding unit 108 correction pattern label input unit 109 correction pattern label Holding unit 110 Correction pattern holding unit 111 Correction pattern holding execution unit 112 Usage pattern label input unit 113 Usage pattern label holding unit 114 Correction pattern reading execution unit 115 Prosody control unit 116 Prosody information holding unit 117 Speech generation unit 118 Speech parameter holding unit 119 DA converter 120 Voice execution unit

Claims (12)

[Claims] 1. A generating unit for performing phonetic language processing on a sentence to be subjected to voice synthesis to generate phonetic data for voice synthesis, and a desired modification to the phonetic data produced by the generating unit. A voice synthesizing apparatus comprising: a correction unit; and a generation unit that generates a parameter of a synthetic voice based on the phonetic data corrected by the correction unit. 2. The correction means includes display means for displaying the phonetic data generated by the generating means, and input means for inputting a desired correction instruction to the phonetic data displayed by the display means. The speech synthesis apparatus according to claim 1, further comprising: 3. The voice synthesizing apparatus according to claim 1, wherein the generating unit generates phonetic data of a sentence for which voice synthesis is performed in units of pause phrases. 4. The phonetic data modified by the modifying means includes at least information for determining a pause length, a vocalization speed, a pitch, a volume, and an accent type of each pause phrase. The speech synthesizer according to claim 3. 5. The storage device further comprises holding means for holding a set of phonetic sound data corrected by the correcting means, wherein the correcting means reads the set of phonetic data held by the holding means and causes the generating means to read the set. The speech synthesizer according to claim 1, further comprising means for correcting the generated phonetic data. 6. The holding means holds a set of phonetic data with a unique label attached, and the correcting means sets the phonetic data set held by the holding means by designating a label. 6. The speech synthesis apparatus according to claim 5, further comprising means for correcting the phonetic sound data generated by the generating means according to a set of the acquired phonetic sound data. 7. A method of controlling a speech synthesizer for synthesizing a voice based on an inputted sentence, wherein speech linguistic processing is performed on a sentence to be speech-synthesized to generate phonetic data for speech synthesis. A generating step for performing a desired correction on the phonetic sound data generated in the generating step, and a generating step for generating a parameter of the synthetic voice based on the phonetic sound data corrected in the correcting step. A control method characterized by the above. 8. The correction step includes a display step of displaying the phonetic sound data generated in the generating step, and an input step of inputting a desired correction instruction to the phonetic sound data displayed in the display step. The control method according to claim 7, further comprising: 9. The control method according to claim 7, wherein the generating step generates phonetic data of a sentence for which voice synthesis is performed in units of pause phrases. 10. The phonetic data modified in the modifying step includes, for each pose phrase, at least information for determining the pose length, vocal speed, pitch, volume, and accent type of the pose phrase. The control method according to claim 9. 11. A holding step of holding a set of phonetic sound data corrected in the correcting step, wherein the correcting step reads out the set of phonetic data held in the holding step to generate in the generating step. The control method according to claim 7, further comprising a step of correcting the generated phonetic data. 12. The holding step holds a phonetic data set with a unique label attached, and the correcting step sets a phonetic data set held in the holding process by designating a label. 12. The control method according to claim 11, further comprising the step of: acquiring the phonetic sound data and modifying the phonetic sound data generated in the generating step by the acquired phonetic sound data set.