JP2016114729A - Text message voicing device, text message voicing method, and text message voicing program - Google Patents

Abstract

Provided is a text message speechization device that can synthesize a plurality of text messages of different creators in an audibly different state even when the same speaker model data is used. A message data acquisition unit acquires message data including a text message and information indicating a message creator. The storage unit 103 stores a setting value of at least one parameter among a text message reading speed, a reading pitch, a reading volume, and a voice output localization in association with the message creator. When the message data acquisition unit 102 acquires the message data, the control unit 101 reads the parameter setting value corresponding to the message creator from the storage unit 103. The voice conversion unit 104 converts the text message into voice data using the speaker model data and parameter setting values. [Selection] Figure 1

Description

  The present invention relates to a text message speechization device, a text message speechization method, and a text message speechization program for speechizing a text message by speech synthesis.

  A text message speech converting apparatus that synthesizes a text message by speech synthesis is used in various apparatuses such as a car navigation apparatus. To make a text message into speech by speech synthesis, speaker model data is required.

  The speaker model data includes a large amount of data because it includes a large number of speech element data such as phonemes, words, and phrases necessary for converting a text message into speech. Also, a great deal of man-hour is required to create one speaker model data.

JP 2010-102163 A

  In some cases, a text message created by each of a plurality of creators is voiced by a text message voice device. In such a case, if the speaker model data used for each of the text messages of a plurality of creators is different, the voices and manners of speaking of the creators are different, and the difference between the creators can be recognized.

  However, as described above, the speaker model data has an enormous amount of data, and it takes a lot of man-hours to create the speaker model data. It is difficult to prepare and memorize it in the text message voicing apparatus.

  In view of such problems, the present invention provides a text message speechization device, a text that can audibly utter a plurality of different text messages in different states even when the same speaker model data is used. An object of the present invention is to provide a message sounding method and a text message sounding program.

  In order to solve the above-described problems of the related art, the present invention provides a message data acquisition unit that acquires message data including a text message and information indicating a message creator who created the text message, and the text message A voice conversion unit for converting into voice data using predetermined speaker model data, a reading speed of the text message, a reading pitch, a reading volume, and a voice output from a plurality of speakers; Storage unit that stores voice conversion reference data including a setting value of at least one parameter of the voice output localization indicating the balance of the message corresponding to the message creator so as to be different for each message creator; When the message data acquisition unit acquires predetermined message data, A setting value of a parameter included in the voice conversion reference data stored corresponding to the message creator of the predetermined message data is read from the storage unit, and the voice conversion unit is included in the predetermined message data. A text message voicing apparatus comprising: a control unit that controls the voice conversion unit so as to convert the text message into voice data using the speaker model data and the set value of the parameter I will provide a.

  In addition, in order to solve the above-described problems of the conventional technology, the present invention acquires predetermined message data including a text message and information indicating a message creator who created the text message, and reads out the text message. Voice conversion reference data including a setting value of at least one parameter of speed, reading pitch, reading volume, and voice output localization indicating a balance when voice is outputted from a plurality of speakers is created as a message. The setting value of the parameter corresponding to the message creator is read from the storage unit stored corresponding to the message creator, and the text message included in the predetermined message data is different for each message creator. Are the predetermined speaker model data and the parameters read from the storage unit. Providing a text message voicing method characterized by converting the audio data by using the set value of over data.

  Furthermore, in order to solve the above-described problems of the prior art, the present invention obtains message data including a text message and information indicating a message creator who created the text message in a computer. Voice conversion reference data including a setting value of at least one parameter of a reading speed, a reading pitch, a reading volume, and a sound output localization indicating a balance when sound is output from a plurality of speakers, The step of reading the set value of the parameter from the storage unit stored in correspondence with the message creator, and the text message included in the message data, as different for each message creator, Model data and the parameters read from the storage unit Providing a text message audio programs, characterized in that and a step of converting the voice data by using the value.

  According to the text message speech device, text message speech method, and text message speech program of the present invention, even if the same speaker model data is used, a plurality of text messages with different creators are aurally different. Can be voiced.

It is a block diagram which shows the text message audio | voice sound apparatus of each embodiment. It is a block diagram which shows the functional internal structure of the control part 101 in the text message speech apparatus of 1st Embodiment. It is a flowchart for demonstrating the operation | movement of the text message audio | voice sound apparatus of 1st Embodiment, the text message audio | voice sound method of 1st Embodiment, and the process by a text message voice sound program. It is a figure which shows an example of the format of message data. It is a figure which shows an example of the audio | voice conversion reference data used by 1st Embodiment. It is a figure for demonstrating the operation | movement of new production | generation of voice conversion reference data, and registration. It is a figure which shows an example of the audio | voice conversion reference data used by 2nd-4th embodiment. It is a flowchart for demonstrating the operation | movement of the text message audio | voice conversion apparatus of 2nd Embodiment, the text message audio conversion method of 2nd Embodiment, and the process by a text message audio conversion program. It is a block diagram which shows the functional internal structure of the control part 101 in the text message speech apparatus of 3rd Embodiment. In 3rd Embodiment, it is a partial flowchart for demonstrating the 1st example which adds the character string which shows the message sender's introduction sentence to a text message. In 3rd Embodiment, it is a partial flowchart for demonstrating the 2nd example which adds the character string which shows the message sender's introduction sentence to a text message. It is a block diagram which shows the functional internal structure of the control part 101 in the text message speechization apparatus of 4th Embodiment.

  In the following, a text message voice conversion device, a text message voice conversion method, and a text message voice conversion program of each embodiment will be described with reference to the accompanying drawings.

<First Embodiment>
In FIG. 1, an external system 200 and an audio output unit 300 are connected to the text message audio device 100. The text message audio device 100 and the external system 200 may be connected via the Internet.

  The external system 200 is a system that manages message data such as an instant messenger, for example.

  The text message speech apparatus 100 includes a control unit 101 that controls the entire text message speech apparatus 100, a message data acquisition unit 102, a storage unit 103, and a voice conversion unit 104 connected to the control unit 101, respectively.

  The message data acquisition unit 102 acquires message data from the external system 200. When the message data acquisition unit 102 requests the external system 200 to transmit the message data, the message data acquisition unit 102 may acquire the message data, or the message data acquisition unit 102 passively receives the message data. You may get it.

  The storage unit 103 stores voice conversion reference data including identification information of speaker model data used for voice conversion of the message data by the voice conversion unit 104 and parameter data to be described later, as a sender (creator) of the message data. Remember every one.

  The voice conversion unit 104 stores at least one speaker model data. The voice conversion reference data stored in the storage unit 103 may be assigned common speaker model data to a plurality of senders.

  As illustrated in FIG. 2, the control unit 101 includes a voice conversion reference data generation unit 1011, a voice conversion reference data writing unit 1012, and a voice conversion reference data reading unit 1013 as functional internal configurations. The voice conversion reference data generation unit 1011, the voice conversion reference data write unit 1012, and the voice conversion reference data read unit 1013 can be configured by software.

  The control unit 101 converts the text message included in the message data acquired by the message data acquisition unit 102 and the voice conversion reference data read from the storage unit 103 by the voice conversion reference data reading unit 1013 into a voice conversion unit in a predetermined format. 104 is supplied.

  The voice conversion unit 104 converts the text message into voice data by using predetermined speaker model data and further referring to the voice conversion reference data.

  The control unit 101 supplies the audio data (audio signal) to the audio output unit 300 and outputs it as audio. The audio output unit 300 includes a speaker and an amplification unit. The audio data is converted into an analog signal by the control unit 101 or a D / A converter (not shown) and supplied to the speaker.

  With reference to the flowchart shown in FIG. 3, the operation of the text message voice generating apparatus of FIG. 1 and the text message voice generating method executed by the text message voice generating apparatus will be further described.

  In FIG. 3, the control unit 101 captures the message data acquired by the message data acquisition unit 102 in step S01. The message data is data as shown in FIG. 4 as an example. The format of the message data is, for example, an XML format.

  As shown in FIG. 4, the message data has a message tag, and the message tag stores various tags related to the message. The account tag stores an account name that can uniquely identify the message sender (message creator). The gender tag stores the gender of the message sender, and the language tag stores the language of the message. The text tag stores a text message that is a message body.

  In step S02, the control unit 101 determines whether or not the voice conversion reference data corresponding to the message sender exists in the storage unit 103.

  FIG. 5 shows an example of the voice conversion reference data stored in the storage unit 103. The voice conversion reference data includes an account name, identification information for specifying speaker model data to be used, and a setting value of parameter data. Here, setting values for the reading speed, reading voice, reading volume, and sound output localization are set as parameter data. The parameter data may be a set value of at least one of these.

  In FIG. 5, the reading speed, the reading voice, and the reading volume are all set to a normal setting value, and the voice output localization is set to a setting value L5: R5. The sound output localization L5: R5 indicates a balance in which sound is output from the left speaker and the right speaker in the two left and right speakers. The sound output localization may be a balance in which sound is output from four speakers, front, rear, left, and right, and may indicate a balance in which sound is output from a plurality of speakers.

  The control unit 101 determines whether there is voice conversion reference data corresponding to the message sender depending on whether voice conversion reference data including the account name described in the account tag included in the message data exists. Can be determined.

  If the voice conversion reference data corresponding to the message sender exists (YES), the control unit 101 shifts the process to step S05, and if not (NO), shifts the process to step S03.

  In step S03, the voice conversion reference data generation unit 1011 generates new voice conversion reference data so that parameter data setting values do not overlap. The speaker model data may overlap with that in other speech conversion reference data.

  When the voice conversion reference data includes a plurality of parameter setting values, the voice conversion reference data generation unit 1011 may generate new voice conversion reference data so that the combination patterns of the parameter data setting values do not overlap. .

  When the voice conversion reference data includes only one parameter setting value, the voice conversion reference data generation unit 1011 generates new voice conversion reference data so that the parameter setting values do not overlap.

  In order to make a plurality of text messages audibly sound differently, it is preferable to greatly change the setting values of the parameter data. Therefore, for example, if the setting values of the reading speed, reading voice, and reading volume are set to three levels of low, normal, and high, respectively, the combination pattern of the setting values of the parameter data is a relatively limited pattern.

  In step S04, the voice conversion reference data writing unit 1012 writes and registers the newly generated voice conversion reference data in the storage unit 103, and shifts the processing to step S05.

  As shown in FIG. 6A, it is assumed that the voice conversion reference data shown in FIG. 5 in which the account name is set to account0 is registered in the storage unit 103 in advance.

  Assume that a message sender having the account name account1 shown in FIG. 4 has sent message data.

  The voice conversion reference data writing unit 1012 registers the voice conversion reference data corresponding to the account name account1 in the storage unit 103 in step S04. Therefore, as shown in FIG. 6B, the storage unit 103 stores the voice conversion reference data corresponding to the existing account name account0 and the voice conversion reference data corresponding to the account name account1. Become.

  As shown in FIG. 6B, for the voice conversion reference data corresponding to the account name account1, for example, the reading speed and the reading pitch are set to high. When the voice conversion reference data includes a plurality of parameter data, the setting value of at least one parameter data may be changed. The speaker model data corresponding to the account name account1 is the same as the speaker model data corresponding to the account name account0.

  By the way, when the voice conversion unit 104 stores, for example, male speaker model data and female speaker model data, the voice conversion reference data generation unit 1011 generates voice conversion reference data. At this time, speaker model data corresponding to the gender described in the gender tag included in the message data may be selected.

  Returning to FIG. 3, in step S <b> 05, the voice conversion reference data reading unit 1013 reads voice conversion reference data corresponding to the message sender from the storage unit 103 and supplies data to be described later to the voice conversion unit 104.

  The format of data supplied from the control unit 101 to the audio conversion unit 104 is, for example, as follows. The control unit 101 uses the text message described in the text tag, the identification information indicating the speaker model data included in the speech conversion reference data, and parameter data other than the speech output localization, for example, data in the SSML format Is generated.

  The control unit 101 supplies SSML format data and audio output localization parameter data to the audio conversion unit 104. The reason why the parameter data for the sound output localization is separated from the data in the SSML format is that the parameter data for the sound output localization cannot be described in the data in the SSML format.

  In step S06, the control unit 101 causes the voice conversion unit 104 to execute voice conversion processing. The voice conversion unit 104 uses the speaker model data of the identification information described in the input SSML format data, and in accordance with the parameter data described in the SSML format data and the parameter data of the voice output localization, the text message Is converted to speech.

  In step S07, the control unit 101 causes the audio output unit 300 to output audio based on the audio data converted by the audio conversion unit 104, and ends the processing.

  The control unit 101 may be configured by a microcomputer, and may be configured to cause the microcomputer to execute the processing shown in FIG. 3 by a computer program (text message speechization program).

  As described above, according to the first embodiment, even when the speaker model data used when the voice conversion unit 104 converts the text message into voice is the same, a plurality of text messages with different message creators can be heard. Can be voiced in different states.

Second Embodiment
The text message sounding apparatus, text message sounding method, and text message sounding program of the second embodiment will be described with a focus on differences from the first embodiment.

  In the second embodiment, as shown in FIG. 7, information on the last use date and time of the voice conversion reference data is added to the voice conversion reference data.

  In the flowchart of the second embodiment shown in FIG. 8, the same steps as those in the flowchart of the first embodiment shown in FIG. In FIG. 8, if there is no voice conversion reference data corresponding to the message sender in the storage unit 103 in step S02 (NO), the control unit 101 shifts the process to step S08.

  In step S08, the control unit 101 determines whether or not a combination pattern of parameter data setting values is insufficient. If there is only one parameter data, it may be determined whether or not the set value of the parameter data is insufficient.

  If the combination pattern is insufficient (YES), the control unit 101 shifts the process to step S09. If the combination pattern is not insufficient (NO), the control unit 101 shifts the process to step S03.

  In step S09, the voice conversion reference data generation unit 1011 deletes the voice conversion reference data having the oldest last use date and time, generates new voice conversion reference data, and shifts the processing to step S05.

  Specifically, the voice conversion reference data generation unit 1011 generates voice conversion reference data in which the account name of the voice conversion reference data with the oldest last use date is replaced with another account name, and the voice conversion reference data writing unit 1012 May be written in the storage unit 103. Along with this, the last use date and time is updated.

  By the process of step S09, the voice conversion reference data set for a message sender having the oldest last use date and time is deleted, and the voice conversion reference data is set for a new message sender. become.

  According to the second embodiment, in addition to the same effects as in the first embodiment, voice conversion reference data can be set for a new message sender even when the combination pattern of parameter data setting values is insufficient. it can.

  As described above, since the combination pattern of the setting values of the parameter data is limited, the second embodiment is effective.

<Third Embodiment>
The text message speechization apparatus, text message speechization method, and text message speechization program of the third embodiment will be described with a focus on differences from the first or second embodiment. The configuration and operation of the following third embodiment may be added to the first embodiment or may be added to the second embodiment.

  As shown in FIG. 9, in the third embodiment, the control unit 101 further includes a character string adding unit 1014 as a functional internal configuration. The character string adding unit 1014 can be configured by software.

  In FIG. 10, the control unit 101 generates new voice conversion reference data in step S03, and registers the voice conversion reference data in the storage unit 103 in step S04. In step S11, the character string adding unit 1014 adds a character string indicating an introduction sentence of the message sender before the text message, and shifts the processing to step S05.

  The storage unit 103 stores templates of introduction sentences for each language. For example, if you as an example the Japanese of the template, in the storage unit 103, as an example, "Hello. It is ???." That the template is stored.

  The control unit 101 reads a language template described in the language tag, and the character string adding unit 1014 adds an account name described in the account tag to the template to generate an introduction sentence of the message sender.

  If the language tag indicates that it is in Japanese, the character string adding unit 1014 generates a character string by replacing the “???” part of the above template with the account name, and uses it as an introduction sentence of the message sender. Append before the text message. The same applies to other languages such as English.

  The character string adding unit 1014 may add an introduction sentence of the message sender before the text message at a timing other than the timing when the new voice conversion reference data is generated. For example, the character string adding unit 1014 adds an introductory sentence before the text message when the voice conversion reference data is used after a predetermined period has elapsed since the last use date and time of the voice conversion reference data.

  When an introductory sentence is added to a text message when a predetermined period or more has passed since the last use date and time of the voice conversion reference data, the voice conversion reference data format is a voice including information on the last use date and time as shown in FIG. This is converted reference data.

  As illustrated in FIG. 11, the control unit 101 determines whether or not a predetermined period or more has elapsed since the last use date and time in step S12 following steps S02 and S04. The control unit 101 shifts the process to step S13 if the predetermined period or more has elapsed (YES), and shifts the process to step S05 if the predetermined period or longer has not elapsed (NO).

  In step S13, the character string adding unit 1014 adds a character string indicating the message sender's introduction before the text message, and proceeds to step S05.

  According to 3rd Embodiment, in addition to the same effect as 1st Embodiment or 2nd Embodiment, there exist the following effects.

  According to the process shown in FIG. 10, which is the first example, when a new message sender sends a text message and the voice conversion unit 104 converts the text message into voice, the message sender's introduction is reproduced. be able to. Therefore, it is possible to recognize who the new message sender is.

  According to the process shown in FIG. 11 as the second example, the voice conversion unit 104 uses the voice conversion reference data in a state in which a predetermined period or more has passed since the last use date and time of the voice conversion reference data to convert the text message into a voice message. When converting, the message sender's introduction can be played. Therefore, even if the message sender is forgotten as the period elapses, it is possible to recognize who the message sender is.

<Fourth embodiment>
The text message sounding apparatus, text message sounding method, and text message sounding program of the fourth embodiment will be described with a focus on differences from the first to third embodiments. The configuration and operation of the following fourth embodiment may be added to each of the first to third embodiments.

  As shown in FIG. 12, in the fourth embodiment, the control unit 101 further includes an audio conversion reference data changing unit 1015 as a functional internal configuration. The voice conversion reference data changing unit 1015 can be configured by software. The character string adding unit 1014 in FIG. 12 can be omitted.

  Also in the fourth embodiment, the format of the voice conversion reference data is voice conversion reference data including information on the last use date and time as shown in FIG.

  First, assume that message sender A has sent a text message. It is assumed that the sound output localization set for the message sender A stored in the storage unit 103 is 3: 7 left and right. At this time, the voice conversion unit 104 converts the text message into voice by setting the voice output localization to 3: 7 left and right.

  Thereafter, it is assumed that the message sender B transmits a text message. It is assumed that the sound output localization set for the message sender B stored in the storage unit 103 is 5: 5 on the left and right. At this time, the voice conversion unit 104 converts the text message into voice with the voice output localization set to 5: 5 on the left and right.

  Furthermore, after that, it is assumed that the message sender B sends a text message again. At this time, the elapsed period from the last use date and time of the voice conversion reference data corresponding to the message sender B is shorter than that of the voice conversion reference data corresponding to the message sender A.

  The voice conversion reference data changing unit 1015 changes the voice output localization set for the message sender B stored in the storage unit 103 to 3: 7 left and right, and the voice conversion unit 104 A text message is converted into a voice by setting the output localization to 3: 7 left and right.

  In addition, the voice conversion reference data changing unit 1015 changes the voice output localization set for the message sender A stored in the storage unit 103 from left / right 3: 7 to left / right 5: 5.

  As described above, the fourth embodiment includes the voice conversion reference data changing unit that changes the setting value of the voice output localization according to the elapsed period from the last use date and time. According to 4th Embodiment, an effect is exhibited, for example in the following cases. Assume that left and right speakers are mounted on a vehicle. Here, it is assumed that the steering wheel is arranged on the right side of the vehicle.

  When the text message transmitted by the message sender A is converted into voice data by the voice conversion unit 104 and the voice output localization is set to 3: 7 left and right and reproduced by the left and right speakers, the driver can move from the position closer to the driver side. You will hear the voice.

  Next, it is assumed that the text message transmitted by the message sender B is converted into voice data by the voice conversion unit 104 and played back by the left and right speakers with the voice output localization set to 5: 5 on the left and right. In this case, since the sound can be heard from almost the center of the left and right, the driver can hear the sound from a position away from the driver as compared with the case where the sound output localization is set to 3: 7 left and right. .

  After that, the text message transmitted again by the message sender B is converted into voice data by the voice conversion unit 104, and the voice output localization is set to 3: 7 left and right and reproduced by the left and right speakers. You will hear the sound from the position you approached.

  Message sender B is sending message data more recently than message sender A. Therefore, the message sender B is more likely to be the center of the topic when the message data is transmitted and received between the message receiver and the message sender than the message sender A.

  The driver listens to the voice of the message sender who is likely to be the center of the topic from a position close to the driver, and listens to the voice of the other message sender from a position relatively far from the driver. Thus, it is possible to easily identify both.

  The present invention is not limited to the embodiments described above, and various modifications can be made without departing from the scope of the present invention. As shown in FIG. 5 and the like, the speech conversion reference data includes identification information for specifying the speaker model data, but if the speech conversion unit 104 has only one speaker model data, the speaker model data The identification information may be omitted.

101 Control unit 102 Message data acquisition unit 103 Storage unit 104 Voice conversion unit
1011 Voice conversion reference data generator
1014 String addition part
1015 Voice conversion reference data change part

Claims (7)

A message data acquisition unit for acquiring message data including a text message and information indicating a message creator who created the text message;
A voice conversion unit that converts the text message into voice data using predetermined speaker model data;
A setting value of at least one parameter of a reading speed of the text message, a reading pitch, a reading volume, and an audio output localization indicating a balance when outputting the sound based on the sound data from a plurality of speakers. A storage unit that stores voice conversion reference data including the corresponding message creator so as to be different for each message creator;
When the message data acquisition unit acquires predetermined message data, the setting value of the parameter included in the voice conversion reference data stored corresponding to the message creator of the predetermined message data is read from the storage unit The voice conversion unit controls the voice conversion unit to convert a text message included in the predetermined message data into voice data using the speaker model data and the set value of the parameter. A control unit,
A text message voicing apparatus comprising: When the message data acquisition unit acquires predetermined message data and the storage unit does not store voice conversion reference data corresponding to the message creator of the predetermined message data,
The controller is
When the voice conversion reference data includes setting values of a plurality of parameters, new voice conversion reference data is generated so that combination patterns of parameter data setting values do not overlap, and the voice conversion reference data includes one parameter. 2. The text message voice generating apparatus according to claim 1, further comprising: a voice conversion reference data generation unit that generates new voice conversion reference data so that parameter setting values do not overlap when only the setting values are included. The voice conversion reference data includes information indicating the last use date and time using the voice conversion reference data,
When the voice conversion reference data generation unit cannot generate new voice conversion reference data in which the combination pattern of parameter data setting values or parameter setting values do not overlap, based on the information indicating the last use date and time, The text message speech conversion apparatus according to claim 2, wherein the voice conversion reference data having the oldest last use date is deleted to generate new voice conversion reference data. The controller is
The character string adding unit for adding a character string, which is an introduction sentence of the message creator, to the message data based on information indicating the message creator. The text message voice converting device described in 1. The voice conversion reference data includes, as the parameter, a setting value of the voice output localization, and information indicating the last use date and time using the voice conversion reference data,
The controller is
The text message voice conversion device according to claim 1, further comprising: a voice conversion reference data changing unit that changes a setting value of the voice output localization according to an elapsed period from the last use date and time. Obtaining predetermined message data including a text message and information indicating a message creator who created the text message;
A voice conversion reference including a setting value of at least one parameter of the reading speed of the text message, a reading pitch, a reading volume, and a voice output localization indicating a balance when voice is outputted from a plurality of speakers. The setting value of the parameter corresponding to the message creator is read out from the storage unit stored corresponding to the message creator so that the data is different for each message creator,
A text message voice, wherein the text message included in the predetermined message data is converted into voice data using predetermined speaker model data and the set value of the parameter read from the storage unit. Method. On the computer,
When message data including a text message and information indicating the message creator who created the text message is acquired, the text message reading speed, reading pitch, reading volume, and voice are output from a plurality of speakers. Voice conversion reference data including a setting value of at least one parameter of voice output localization indicating a balance at the time of output is stored in association with the message creator so as to be different for each message creator. Reading the setting value of the parameter from the storage unit;
Converting the text message included in the message data into speech data using predetermined speaker model data and the setting values of the parameters read from the storage unit;
A text message voicing program characterized in that

Images