JP2018159777A - Voice reproduction device, and voice reproduction program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make it possible to perform abundant expression of intension, suppressing increment of the storage capacity required for storing voice data of answers.SOLUTION: The voice reproduction device includes an answer acquiring unit which acquires voice data of an answer to a question represented by an input voice signal, an intention specifying unit which designates an intention to be given to the answer, a rhythm control data acquisition unit which acquires rhythm control data representing a time change of the rhythm according to the intention designated by the intention specifying unit, and an answer reproduction unit which reproduces a voice of an answer in which a time change of rhythm based on the voice data is controlled according to the rhythm control data.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an audio reproduction technique.

  As an application example of the sound reproduction technology, there is a sound interaction between a person and a machine or a sound interaction between a machine and a machine. As an example of a voice interaction between a person and a machine, there is a voice dialogue system that synthesizes and plays back a voice of an answer to a question of a user by voice. An example of voice interaction between machines is that the voice playback device B recognizes the voice of the question played back by the voice playback device A according to a predetermined scenario and plays back the voice of the answer, Specifically, there are plays and comics where all the characters are played by a machine (voice playback device). When synthesizing the voice of the answer to the question by voice, it is preferable to reproduce the voice of the answer with the intention in order to realize a human-like answer to the question by the voice of the user. For example, in the technique disclosed in Patent Document 1, in order to express the intention included in the answer, the pitch of the ending is made different between a positive answer and a negative answer.

Japanese Patent Application Laid-Open No. 2015-064480

  However, various intents cannot be expressed only by adjusting the ending pitch of the answer as in the technique disclosed in Patent Document 1. If voice data of answers is prepared for each intention for each of various intentions, a rich intention can be expressed, but there is a problem that the storage capacity of the storage device for storing the voice data increases.

  The present invention has been made in view of the problems described above, and it is an object of the present invention to provide a technique capable of expressing a rich intention while suppressing an increase in storage capacity required for storing voice data of answers. And

  In order to achieve the above object, an audio reproduction device according to an aspect of the present invention specifies an answer acquisition unit that acquires audio data of an answer to a question represented by an input audio signal, and an intention to be given to the answer An intent specifying unit, a prosody control data acquiring unit that acquires prosody control data representing a temporal change in prosody according to the intention specified by the intention specifying unit, and a temporal change in prosody based on the speech data. An answer reproducing unit that reproduces the sound of the answer controlled according to the above.

  The time change of the prosody means a time change of each prosody component such as pitch, speech speed, volume, and speech timing. Note that the speech speed refers to the number of phonemes that are pronounced per unit time. When a person utters an answer to a question, he adjusts the time change of the prosody according to the intention to be included in the answer, so that various types such as `` easy '' and `` careful '' or `` anger '' and `` are '' Express intentions. According to this aspect, it is possible to reproduce the answer voice in which the time change of the prosody is finely controlled according to the intention to be included in the answer, and it is possible to express a rich intention. Here, the prosodic control data is an array of changes in prosodic components at each time (time starting from the beginning of the answer) of prosodic components such as pitch, speech speed, volume, and utterance timing, that is, sequence data. The amount of data is small compared to the waveform data of speech. For this reason, even if prosodic control data corresponding to each intention is stored in the storage device in advance for each answer, it is compared with a mode in which the sound data of each answer sounded with a different intention is stored in the storage device. Therefore, it can be handled with a small storage capacity. That is, according to this aspect, it is possible to express a rich intention while suppressing an increase in storage capacity required for storing the answer voice data. Note that the answer is not limited to a specific answer to the question, but includes an answer (interjection). Answers include answers to questions and dialogues, as well as dialogues in theater and comics. In addition to human voices, animals such as “one” (bowwow) and “meow” (meow) A cry is also included. That is, the answer and the voice here are concepts including not only a voice uttered by a person but also an animal cry.

  In a more preferred aspect, the prosody control data has a database in which each intention is stored, and the prosody control data acquisition unit acquires prosody control data corresponding to the intention designated by the intention designation unit from the database. To do. In another preferred aspect, the prosodic control data acquisition unit, when prosodic control data corresponding to the intention specified by the intention specifying unit is not stored in the database, It is obtained by interpolation using a plurality of prosodic control data stored in a database. According to such an aspect, it is possible to express more intentionally while suppressing an increase in the storage capacity of the database. In another preferred aspect, the system further comprises an intention specifying unit that analyzes the input audio signal and specifies an intention given to the question, and the intention specifying unit is specified by the intention specifying unit. The intention to be given to the answer is designated according to the intended intention. According to such an aspect, it becomes possible to reproduce the voice of an answer with an intention corresponding to the intention included in the question.

  The aspect of the present invention can be conceptualized as a program that causes a computer to function as the sound reproducing device as well as the sound reproducing device.

It is a block diagram which shows the structure of the audio | voice reproduction apparatus which concerns on embodiment. It is a figure which shows an example of the time change of the prosody in the answer with an intention. It is a flowchart which shows operation | movement of an audio | voice reproduction apparatus.

Embodiments of the present invention will be described below with reference to the drawings.
(A: Configuration)
FIG. 1 is a diagram showing a configuration of an audio playback device 10 according to an embodiment of the present invention.
The audio reproduction device 10 is, for example, a device incorporated in a stuffed animal. When the user asks the stuffed animal, the voice playback device 10 synthesizes and plays back the voice of the answer with the intention specified by the user. When a person utters an answer to a question, he adjusts the time change of the prosody according to the intention included in the answer. Express intentions. For example, FIG. 2 shows a time waveform TW of a reference voice “Anasa” that is pronounced without a specific intention, and a prosody of a reference voice of “Anasa” that includes “comfort”. The change pattern P1 and the prosody change pattern P2 from the reference voice of the voice “Anasa” with “carefulness” are exemplified. In FIG. 2, the amount of change from the reference speech of each prosodic component of “pitch”, “speech speed”, and “volume” is represented by the position on the coordinate axis from the center of gravity of the triangle to each vertex. This means that the further away from the center of gravity, the higher the pitch, the faster the speech speed, and the higher the volume compared to the reference voice.

  The audio reproduction device 10 includes a CPU (Central Processing Unit), an audio input unit 102, and a speaker 114. When the CPU executes an application program installed in advance, a plurality of functional blocks are constructed as follows. Is done. Specifically, in the voice reproduction device 10, a language analysis unit 104, an answer acquisition unit 106, an intention designation unit 108, a prosody control data acquisition unit 110, and an answer reproduction unit 112 are constructed. Although not particularly illustrated, the sound reproducing device 10 also includes a display unit, an operation input unit, and the like. The user can check the state of the device and input various operations to the device. Various settings can be made. The audio playback device 10 is not limited to a toy such as a stuffed toy, and may be a so-called pet robot, a terminal device such as a mobile phone, a tablet personal computer, or the like.

  Although not described in detail, the sound input unit 102 includes a microphone that converts sound into an electrical signal and an A / D converter that converts the converted sound signal into a digital signal. The language analysis unit 104 analyzes the semantic content of the question defined by the voice signal input from the voice input unit 102, and obtains the semantic content data indicating the analysis result (that is, the semantic content of the question). To give.

  The answer library 124 is a database that stores in advance a plurality of pairs of identifiers (hereinafter referred to as answer identifiers) that uniquely indicate answers to questions by users and voice data of the answers. This voice data is a recording of the voice of a model person. For example, “Yes”, “No”, “Yes”, “Yes”, “Fun”, “I see” For example, it is a companion. The answer voice data is in a format such as wav or mp3, for example. Specific examples of answer identifiers include character strings and serial numbers representing the contents of answers such as “Yes” and “No”. Note that the answer to the question is not limited to answering or comprehension, for example, a sentence that presents the information requested by the question, such as “Sunny” for the question “What is the weather today?” May be.

  The answer acquisition unit 106 selects one answer identifier and voice data of an answer to the question whose meaning is represented by the meaning contents data given from the language analysis unit 104 from the answer library 124, and selects the selected answer identifier and Voice data is read from the answer library 124 and acquired. Then, the response acquisition unit 106 outputs the response identifier acquired from the response library 124 to the prosodic control data acquisition unit 110, and outputs the audio data acquired from the response library 124 to the response playback unit 112. The answer acquisition unit 106 according to the present embodiment includes a table that stores answer identifiers of answers that are appropriate to the question of the meaning content represented by the meaning content data in association with the meaning content data that can be output from the language analysis unit 104. ing. The answer acquisition unit 106 refers to the contents stored in the table and selects one answer that is appropriate for the question of the meaning content represented by the meaning content data.

  In the present embodiment, an aspect in which an answer is selected according to the meaning content of the question will be described, but an answer may be selected at random regardless of the meaning content of the question. In this case, the language analysis unit 104 and the table are not necessary. Specifically, the output signal of the voice input unit 102 is given to the answer acquisition unit 106, and the answer acquisition unit 106 receives an answer identifier and voice data from the answer library 124 when receiving the voice signal from the voice input unit 102. It is sufficient to execute a process of randomly reading out.

  The prosodic library 122 is associated with each of a plurality of response identifiers stored in the response library 124, and an identifier indicating the intention (hereinafter referred to as an intention identifier) for each intention included in the response indicated by the response identifier. And a prosody control data for defining a temporal change of the prosody in the corresponding answer. Here, the prosody control data is an array of changes in prosodic components such as pitch, speech speed, and volume at each time (time starting from the beginning of the answer), that is, sequence data. As a specific example of the intention identifier, there is a character string representing the content of the intention such as “anger” or “drill”.

  Prosodic control data can be created in the following manner. For example, if the answer is “Anosa”, the waveform data of the voice “Anosa” that is pronounced on a flat plate without any specific intention is used as the reference data, and a specific intention such as “Angry” or “Akure” Compare the waveform data of the voice “Anosa” that is pronounced with the above reference data, and calculate the difference (offset) from the reference data at each time for each prosodic component such as pitch, speech speed, and volume. It is conceivable that the prosody control data is calculated and arranged for each component in order of time. In this case, the answer library 124 may store the sound data of the sound generated on the flat plate without any specific intention as the sound data of the answer. Note that the prosody control data for voices sounded with other intentions may be generated using the waveform data of the voices sounded with a specific intention as reference data. As a result, the voice data of the voice generated with the specific intention may be stored in the answer library 124.

  The intention designating unit 108 is a device that allows the user to designate one of the intentions whose intention identifiers and prosodic control data are stored in the prosody library 122 as intentions that can be included in the answer to the question. For example, the intention designating unit 108 displays a list of intention identifiers on a display unit (not shown in FIG. 1), and displays the intention identifier designated by an operation on the operation input unit (not shown in FIG. 1). 110.

  The prosodic control data acquisition unit 110 reads out and acquires the prosodic control data corresponding to the intention identifier given from the intention designating unit 108 and the answer identifier given from the answer obtaining unit 106 from the prosody library 122. The prosody control data acquisition unit 110 gives the prosody control data read from the prosody library 122 to the answer reproduction unit 112.

The answer reproducing unit 112 converts the voice represented by the voice data given from the answer obtaining unit 106 according to the prosodic control data given from the prosody control data obtaining unit 110 to the time of each prosodic component of pitch, speech speed, and volume. The changes are controlled separately and reproduced (synthesized).
The above is the configuration of the audio reproduction device 10.

(B: Operation)
Next, the operation of the audio playback device 10 will be described.
FIG. 3 is a flowchart showing the processing operation in the audio playback device 10. In the present embodiment, when the user makes an inquiry by voice to the stuffed animal to which the audio reproduction device 10 is applied, the processing shown in this flowchart is started.

  When the user makes a question by voice, the voice of the question is converted into a voice signal by the voice input unit 102, and the voice signal is supplied from the voice input unit 102 to the language analysis unit 104. In step Sa <b> 11, the language analysis unit 104 accumulates the audio signal supplied from the audio input unit 102 in a memory or the like, and determines whether or not the voice inquiry has ended. Whether or not the inquiry has ended is determined by whether or not the state in which the volume of the audio signal has become less than a predetermined threshold has continued for a predetermined time. If the question has not ended (if the determination result in step Sa11 is “No”), the language analysis unit 104 executes the process in step Sa11 again and waits for the end of the question utterance.

  In step Sa12 which is executed when the determination result in step Sa11 is “Yes”, the language analysis unit 104 analyzes the semantic content of the question defined by the voice signal accumulated in the memory or the like, and indicates the semantic content indicating the analysis result. Data is given to the answer acquisition unit 106. In the present embodiment, the determination as to whether or not the utterance of the question has ended and the semantic analysis of the question are executed sequentially, but both may be executed sequentially in parallel. In this way, the delay from the end of the question utterance to the completion of the semantic analysis can be reduced. In step Sa13 subsequent to step Sa12, the answer acquisition unit 106 acquires the answer identifier and voice data of the answer to the question whose meaning is represented by the meaning contents data from the answer library 124, and the former is the prosody control data acquisition unit. 110 and the latter is given to the answer reproducing unit 112.

  In step Sa14 subsequent to step Sa13, the answer reproducing unit 112 determines whether or not the answer is being reproduced. The case where the answer reproducing unit 112 determines that the answer is being reproduced in step Sa14 is a case where the user asks the next question while reproducing the answer in response to a certain question. If the answer is being reproduced (if the determination result in step Sa14 is “Yes”), the answer reproduction unit 112 executes the process in step Sa14 again until the determination result in step Sa14 becomes “No”. When the determination result in step Sa14 is “No”, the intention specifying unit 108 causes the user to specify the intention that can be included in the answer, and outputs an intention identifier indicating the intention specified by the user to the prosodic control data acquiring unit 110. (Step Sa15). In step Sa16 subsequent to step Sa15, the prosody control data acquisition unit 110 acquires, from the prosody library 122, prosodic control data corresponding to the intention identifier given from the intention specifying unit 108 and the answer identifier given from the answer acquisition unit 106. Then, the acquired prosodic control data is notified to the answer reproducing unit 112 and the reproduction of the voice data of the answer selected by the answer acquiring unit 106 is instructed. In accordance with this instruction, the answer reproducing unit 112 controls the time change of each prosodic component of pitch, speech speed, and volume independently according to the prosodic control data given from the prosodic control data acquisition unit 110, while The voice represented by the voice data is reproduced (step Sa17).

  For example, if the answer “Anosa” is selected in step Sa13 and the intention of “easy” is specified in step Sa15, the prosody changes with time in the time change pattern P1 of FIG. 2 in step Sa17. A voice is synthesized, and a user who listens to this voice feels “comfort” from the time change of the prosody in the voice. On the other hand, if the answer “Anosa” is selected in step Sa13 and the intention of “careful” is designated in step Sa15, the prosody changes with time in the time change pattern P2 of FIG. 2 in step Sa17. A voice is synthesized, and a user who listens to this voice feels "carefulness" from the time change of the prosody in the voice.

  As described above, according to the present embodiment, it is possible to reproduce the answer voice in which the time change of the prosody is finely controlled according to the intention included in the answer, and it is possible to express a rich intention. In the audio reproduction device 10 of this embodiment, it is necessary to store the prosody library 122 in addition to the answer library 124 in a storage device or the like. However, since the prosodic control data stored in the prosodic library 122 is sequence data, the total amount of data is smaller than the sum of speech data with various intentions for each answer. For this reason, an increase in the storage capacity of the storage device can be suppressed as compared with a mode in which all of these audio data is stored. That is, according to the present embodiment, it is possible to express a rich intention while suppressing an increase in the storage capacity of a storage device that stores data required for synthesizing the answer speech. In the above embodiment, the case of controlling the temporal change of the pitch, the speech speed, and the volume has been described. However, the speech timing may be further controlled.

(C: Deformation and application examples)
The present invention is not limited to the above-described embodiments, and various applications and modifications as described below are possible, for example. In addition, one or more arbitrarily selected aspects of application / deformation described below can be appropriately combined.

<Voice input part>
In the above-described embodiment, the voice input unit 102 is configured to input a user's voice (speech) with a microphone and convert the voice into a voice signal, but is not limited to this configuration. That is, the voice input unit 102 may be configured to input or input a speech by a voice signal in some form. Specifically, the voice input unit 102 is configured to input a voice signal processed by another processing unit or a voice signal supplied (or transferred) from another device, and further, is built in an LSI and is simply a voice. This is a concept including an input interface circuit for receiving a signal and transferring it to a subsequent stage. In addition, the question to the voice reproduction device 10 is not limited to the question by voice, and may be input of text data representing a sentence in which the question is written. In this case, instead of the voice input unit 102, a text data input unit may be provided in the voice reproduction device 10.

<Prosodic control unit>
In the above embodiment, prosodic control data is prepared in units of voice data of answers to questions, and the temporal change of the prosody of the answer voice is controlled in units of voice data. However, the prosodic control data is a phoneme unit constituting the answer voice, a waveform sample unit when the answer voice is sampled to generate voice data, or a frame unit when the answer voice is divided into frames of a predetermined time. Prosodic control data may be prepared, and the time change of the prosody may be controlled in units of these.

<Variation of intention designation>
In the above embodiment, the user specifies only one intention that can be included in the answer to the question. For example, the user can specify a plurality of intentions such as “Are” and “Easy”. May be. In this case, the intention specifying unit 108 causes the prosody control data acquiring unit 110 to output an intention identifier indicating each of a plurality of intentions specified by the user. In this way, one prosodic control data that directly corresponds to a plurality of intentions designated by the user is not stored in the prosodic library 122. Therefore, the prosody control data acquisition unit 110 acquires prosody control data corresponding to each of the plurality of intention identifiers, and generates one prosody control data by interpolation using the plurality of prosody control data. It ’s fine. Specifically, the prosody control data acquisition unit 110 generates new prosody control data by weighting and adding the plurality of prosody control data acquired in the above manner, for example, 1: 1, and the like. The data is output to the answer playback unit 112 to synthesize the answer voice.

  According to such an aspect, it is considered that the voice of an answer corresponding to an intermediate intention among a plurality of intentions designated by the user can be synthesized. In general, even if the waveform data of the voice “Ane” that is pronounced “feeling” and the waveform data of the sound “Ane” that is pronounced “feeling easy” are weighted and added, This is because the waveform data of a voice that feels “bright” is not obtained, but the prosody control data is sequence data, so that prosody control data representing intermediate intentions can be generated by weighted addition. Similarly, sounding on a flat plate without a specific intention is also considered an aspect of the intention, and by specifying “flat plate” and “careful”, “flat plate” such as “slightly careful” and “careful” Prosodic control data corresponding to an intermediate intention can also be generated.

  In an aspect in which a user designates a plurality of intentions, the weight of each intention in the above weighted addition may be designated by the user, and the weight for each prosodic component may be designated by the user. This is because the importance of each prosodic component in directing the intention may differ between “anger” and “clear”. Also, let the user specify an intermediate intention among multiple intentions, such as “feeling relaxed with ease”, and decompose the intermediate intention into multiple intentions to generate prosodic control data corresponding to each intention. The prosody control data acquisition unit 110 may acquire the prosody control data. In short, if prosodic control data corresponding to the intention designated by the intention designating unit 108 is not stored in the prosodic library 122 (database), interpolation using a plurality of prosodic control data stored in the database is performed. As long as the prosody control data corresponding to the specified intention is acquired by the prosody control data acquisition unit 110, any method may be used.

  In the above-described embodiment, the user specifies the intention to be included in the answer to the question for each question to the voice playback device 10, but the intention to be included in the answer so that the voice of the answer with the same intention is always synthesized. May be specified in advance by the user. Further, when the answers are synthesized according to a predetermined scenario, the intention may be designated according to the progress of the scenario. Specifically, scenario data in which intention identifiers of intentions that can be included in an answer are arranged in the order of reproduction of the answers is input to the intention specifying unit 108, and each time a questioned voice is input from the voice input unit 102, the intention specifying unit 108 The intention identifiers included in the scenario data may be acquired in the order of description and output to the prosodic control data acquisition unit 110.

  In addition, the personality, atmosphere, and speaking characteristics of the voice reproduction device 10 or an answer character that personifies the stuffed stuffed toy of the voice reproduction device 10 such as “always feel comfortable” or “always keep careful” , Character characteristics) are prepared in advance, each character characteristic is associated with an intention identifier corresponding to the character characteristic, and the user is allowed to specify the character characteristic of the answer character, thereby specifying the intention to be included in the answer You may make it let it. Alternatively, a plurality of answer characters each having different character characteristics may be prepared, and the user may designate the intention to be included in the answer by designating the answer character. When combining the answers to the questions according to the scenario, the character characteristics may be determined for each character in the scenario. In addition, a plurality of character characteristics are associated with one answer character, and it is determined in the scenario which character characteristics are used to synthesize the answer speech. You can produce different usages.

  In addition, an intention specifying unit that analyzes the voice signal of the question and identifies the intention given to the question is provided, and the intention specifying unit 108 responds to the above question according to the intention specified by the intention specifying unit. The intention to give may be specified. Specifically, for each question for which answer voice data is stored in the answer library 124, waveform data of a voice that is pronounced on a flat plate without any particular intention is stored in the intention specifying unit in advance as reference data. In addition, the intention specifying unit converts the speech signal input from the speech input unit 102 into waveform data and compares it with corresponding reference data, so that the prosody definition data representing the prosody in the speech represented by the waveform data. (Data having the same format as the prosodic control data) is generated. Then, the intention specifying unit reads the intention identifier corresponding to the prosody control data that is the same as or close to the prosody definition data from the prosody library 122 and gives it to the intention designating unit 108. The intention designating unit 108 may cause the prosodic control data acquisition unit 110 to output an intention identifier having a predetermined specific relationship with the intention identifier given from the intention specifying unit. Here, with respect to the above-mentioned specific relationship, for example, an opposite intention relationship such as “careful” with respect to “comfort” or the same intention relationship can be considered. When the “same intention relationship” is adopted as the specific relationship, the intention specifying unit 108 should be caused to execute the process of outputting the intention identifier given from the intention specifying unit to the prosodic control data acquiring unit 110 as it is. It ’s fine.

<Evaluation and scoring of how to put intent for questions>
In an aspect in which the intention specifying unit is provided, the intention specified by the intention specifying unit and the stored contents of the prosody library 122 may be compared to evaluate and score how the intention is included in the question. Specifically, regarding the question spoken by the user with intentions such as “anger” and “clear”, whether or not the intention specified by the user was specified by the intention specifying unit, or the intention specification According to the comparison result between the prosody definition data generated from the questioned speech signal by the section and the prosody control data stored in the prosody library 122 for the intention, the suitability of the intention is evaluated and scored. Just do it. According to this aspect, it becomes possible to support practice of how to put in the intention to the question.

<Others>
In the above embodiment, the language analysis unit 104, the prosody library, and the answer library, which are configured to acquire answers to the utterances, are provided on the side of the audio playback device 10, but the processing load is heavy in the terminal device or the like. In view of the above, the storage capacity is limited, and the like may be provided on the external server side. Further, in the above embodiment, the application example of the present invention to the voice reproduction apparatus that synthesizes and reproduces the voice of the answer to the question by the voice of the user to realize the voice dialogue has been described. The present invention may be applied to a sound reproducing device that performs the following.

DESCRIPTION OF SYMBOLS 102 ... Voice input part 104 ... Language analysis part 106 ... Answer acquisition part 108 ... Intention designation part 110 ... Prosody control data acquisition part 112 ... Answer reproduction part 122 ... Prosody library 124 ... Answer library

Claims (5)

An answer acquisition unit for acquiring voice data of an answer to a question represented by the input voice signal;
An intention designating unit for designating an intention to be given to the answer;
A prosody control data acquisition unit for acquiring prosody control data representing a temporal change of the prosody according to the intention designated by the intention designation unit;
An answer playback unit for playing back the answer voice in which the time change of the prosody based on the voice data is controlled according to the prosody control data;
An audio reproducing apparatus comprising: A database in which the prosodic control data is stored for each intention;
The audio reproduction device according to claim 1, wherein the prosodic control data acquisition unit acquires prosodic control data corresponding to the intention specified by the intention specifying unit from the database. The prosodic control data acquisition unit, when prosodic control data corresponding to the intention specified by the intention specifying unit is not stored in the database, the prosodic control data is stored in the database The sound reproducing device according to claim 2, wherein the sound reproducing device is obtained by interpolation using control data. Having an intention specifying unit for analyzing the input voice signal and specifying the intention given to the question;
The audio reproducing apparatus according to claim 1, wherein the intention specifying unit specifies an intention to be given to the answer according to the intention specified by the intention specifying unit. Computer
An answer acquisition unit for acquiring voice data of an answer to a question represented by the input voice signal;
An intention designating unit for designating an intention to be given to the answer;
A prosody control data acquisition unit for acquiring prosody control data representing a temporal change of the prosody according to the intention designated by the intention designation unit;
An answer playback unit for playing back the answer voice in which the time change of the prosody based on the voice data is controlled according to the prosody control data;
A voice reproduction program characterized by being made to function.

Images