JP2004094085A - Speech data distribution apparatus and client terminal device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a reliable speech data distribution apparatus. <P>SOLUTION: The speech data distribution apparatus for distributing speech data through a data network receives at least a series of text data from at least one client terminal, converts the text data by speech synthesis to a series of speech data and accumulates the data for the purpose of delivery. The system corrects and updates the series of accumulated speech data according to reception of a correction request from the client terminal. The client terminal sends the text data to be converted to the series of speech data to a speech data distribution server in response to such speech data distribution apparatus and sends the correction request for correcting and updating the series of speech data to the speech data distribution server. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an audio data distribution device that distributes audio data to user terminals via a data network such as the Internet, and a distribution requester terminal.
[0002]
[Prior art]
When reading an e-mail or a long text on a Web page on a screen, there is a need to present the text in voice to reduce the burden on the eyes. To meet this need, there has been known a method of realizing voice presentation using a voice data distribution device. In such an audio data distribution device, converting document data in a text format or the like into audio data requires special software called an audio synthesis engine. There is also known an audio data distribution system for providing services to users in advance. In this regard, Japanese Patent Application Laid-Open No. 2001-282268 discloses that text data such as a weather forecast is received from a requester, voice synthesis data is created and distributed to a Web server or a telephone server that is a distribution source, so that a homepage viewer or a telephone An audio data distribution system that allows a user to listen to information such as a weather forecast by voice is disclosed. With such a configuration, a distribution requester who wants to distribute audio data can transmit such audio data only by transmitting document data in a text format or the like to the data distribution system without having a voice synthesis function in its own device. It can be provided to users at the destination.
[0003]
By the way, on the premise that Japanese is the target language, speech data is usually synthesized from text data, which is a mixture of kanji and kana characters, by using word dictionary data, prosodic rule data, and speech unit data. Is generated.
[0004]
[Problems to be solved by the invention]
In the voice data generated in this manner, a reading error or an unknown word is always generated, that is, the corresponding voice cannot be determined because the word is not registered in the word dictionary, and the reading is undecided and distributed. It is expected that the danger will occur. In Japanese, in particular, there are problems such as differences in appropriate readings or the generation of many unknown words due to technical terms, depending on the field to which the contents of the text belong, for example, fields such as economy, culture, politics, and entertainment. is expected. As a result, there is a danger that, as a result, voice data containing erroneous reading is delivered, and a user who listens to the voice by voice may misunderstand the information. The provision of such erroneous audio data has impaired the reliability of the audio data distribution service and has been a factor hindering its spread.
[0005]
The present invention has been made in view of the above problems, and an object of the present invention is to provide a reliable audio data distribution device.
[0006]
[Means for Solving the Problems]
An audio data distribution device according to the present invention is an audio data distribution device that distributes audio data via a data network, receives at least a series of text data from at least one client terminal, and converts the text data into one by voice synthesis. Voice data storage means for converting to a series of voice data and storing for distribution, and voice data correction for correcting and updating the stored voice data in response to a correction request from the client terminal Updating means.
[0007]
The distribution request terminal according to the present invention is a requester terminal for requesting a voice data distribution device that distributes voice data via a data network to distribute at least one of the plurality of voice data. Text data transmitting means for transmitting text data to be converted to a series of voice data to the voice data distribution server, and transmitting a correction request for correcting and updating the series of voice data to the voice data distribution server And a correction information transmitting unit.
[0008]
BEST MODE FOR CARRYING OUT THE INVENTION
Embodiments of the present invention will be described in detail with reference to the accompanying drawings.
<First embodiment>
FIG. 1 shows a first embodiment of the present invention, and shows an entire system configuration including an audio data distribution device. The audio data distribution device (also referred to as an audio data distribution system) includes an audio data distribution server 10 and an audio data synthesis server 20.
[0009]
Referring to the voice data synthesis server 20 shown on the right side of FIG. 1, the voice data synthesis server 20 is connected to the Internet 30 and enables communication with the client terminal 40 and / or the user terminal 100. The hardware configuration of the voice data synthesis server 20 is configured as a normal server computer. As shown, the software configuration of the voice data synthesis server 20 includes a control program 22, a TCP / IP unit 23, an HTTP / CGI unit 24, a voice synthesis processing unit 25, a voice data accumulation request unit 26, including. The control program 22 is an operating system that controls the basic control of the entire voice data synthesis server 20. The TCP / IP unit 23 executes a procedure of the TCP / IP protocol under the control of the control program 22, and realizes data communication with the client terminal 40 and / or the user terminal 100 via the Internet 30. The HTTP / CGI unit 24 distributes a Web page for receiving a voice data distribution request to the Internet 30 via the TCP / IP unit 23, and receives text data from the client terminal 40 by using the CGI function. Has functions. The HTTP / CGI unit 24 also has a function of returning the synthesized voice data to the client terminal 40 and receiving a correction instruction. The HTTP / CGI unit 24 implements data communication by HTTP (HyperText Transfer Protocol; IETF: see RFC2616).
[0010]
The speech synthesis processing unit 25 has a function of converting text data input from the HTTP / CGI unit 24 into speech data. The format of the input text data is usually plain text, that is, a JIS kanji character code string, but may be another document format. The output audio data is composed of phoneme data corresponding to the audio waveform, and is usually collected in one file. This file only needs to be a file in an audio waveform format that can be reproduced by the user terminal 100 and / or the client terminal 40, and various file formats such as a WAV file and an MP3 file are assumed. The speech synthesis processing unit 25 includes, for example, a word dictionary data file 251 in which information of notation, reading, accent and part of speech of about 100,000 words is stored in association with each word, and control rule data of fundamental frequency and sound pressure. The stored prosody rule data file 252 and the speech unit data file 253 storing waveform data for each phoneme are connected. In the speech synthesis processing procedure of the speech synthesis processing unit 25, first, an intermediate language, which is a kana character string with accent symbols, is generated from text data that is a sentence mixed with Chinese characters and kana using the word dictionary data of the word dictionary data file 251. Next, from the generated intermediate language, a synthesis parameter including a speech unit number, pitch pattern information, and phoneme duration information is generated using the prosody rule data of the prosody rule data file 252. Then, in accordance with the generated synthesis parameters, the speech unit data of the speech unit data 253 is read out and sequentially connected to generate speech waveform data. Thereby, the audio waveform information of the audio data is generated as, for example, PCM data having a sampling frequency of 8 kHz and quantization of 16 bits, and is temporarily stored as an audio data 63 in an appropriate primary storage device of the audio data synthesis server 20. .
[0011]
The speech synthesis processing unit 25 further has a function of returning an unknown word as a character string. Here, the unknown word is a word that is not registered in the word dictionary data file and may be misread. The speech synthesis processing unit 25 converts a word determined to be an unknown word into a voice with predetermined appropriate reading and outputs a list of words determined to be an unknown word as an unknown word list 64. The unknown word list 64 is temporarily stored in an appropriate primary storage device of the voice data synthesis server 20 together with the previous voice data 63.
[0012]
The audio data storage requesting unit 26 transmits the temporarily stored audio data 63 to the audio data distribution server 10 in response to the OK message of the distribution approval for the audio data from the client terminal 40, and stores the audio data 63 therein. It has a function of requesting distribution to each user terminal 100 toward the Internet 30 at least.
Referring to the audio data distribution server 10 shown on the left side of FIG. 1, the audio data distribution server 10 is connected to the Internet 30 to enable data communication with the user terminal 100 and / or the client terminal 40. . The hardware configuration of the audio data distribution server 10 is configured as a normal server computer. As shown in the figure, the software configuration of the audio data synthesizing server 20 includes a control program 12, a TCP / IP unit 13, an HTTP unit 14, an audio data storage unit 15, and an audio data database 60 for distribution. Hereinafter, referred to as audio data DB).
[0013]
The control program 12 is an operating system that controls the basic control of the entire audio data distribution server device 10. The TCP / IP unit 13 executes a procedure of the TCP / IP protocol under the control of the control program 12, and realizes data communication with the client terminal 40 and / or the user terminal 100 via the Internet 30. The HTTP unit 14 has a function of distributing desired audio data to the user terminal 100 or the requester terminal 40 that requests distribution of the audio data stored in the audio data DB 60 via the TCP / IP unit 13. Have. The audio data DB 60 includes an identifier group including a plurality of identifiers, and an audio data group including a plurality of audio data associated with each of the identifiers. The configuration of the audio data DB 60 can be realized by a normal file system using an identifier as a file name.
[0014]
FIG. 2 shows the internal configuration of the client terminal 40 shown in FIG. Here, the client terminal 40 is connected to the Internet 30 to enable data communication. The hardware configuration of the client terminal 40 is a normal network client terminal such as a personal computer, and a normal display 51 and a keyboard 54 are connected. The hardware configuration of the client terminal 40 further includes a speaker 52 for reproducing and verifying audio data and a microphone 53 for directly creating audio data based on an operator's voice input. The software configuration of the client terminal 40 includes a control program 41 that controls the basic control of the entire client terminal 40, a TCP / IP unit 42 for implementing data communication via the Internet 30, and a keyboard 54. An I / O control unit 43 for realizing input / output of an I / O device in cooperation with a control program 41, an audio input unit 46 for digitally processing and inputting audio input from a microphone 53, and a TCP / IP unit 42. A Web browser unit 44 for displaying a Web page received via the Web browser and transmitting an operator's input to the Web page again to the Internet, and a speaker for outputting audio data captured through the Web page displayed on the Web browser unit 44 And an audio data reproducing unit 45 for reproducing through the audio source 52.
[0015]
Note that the requester terminal 40 uses the microphone 53 and the voice input unit 46 to directly link the voice of the operator with the text input field displayed on the web browser unit 44 for the correction of the text data, and to output the voice data. It is also possible to generate them directly.
FIG. 3 shows an example of a voice data distribution request screen on the client terminal 40. The voice data distribution request screen 80 includes a text data input field 81, a voice data conversion instruction button 82, and a voice synthesis processing condition setting field 83. The text data input field 81 is a field in which the operator of the client terminal 40 inputs a character as a text data of a document that is actually desired to be delivered as voice data. The text data is transmitted to the voice data synthesizing server 20 when the voice data conversion instruction button 82 is specified after the text data is input. The voice synthesis processing condition setting column 83 allows the voice data synthesis server 20 to set detailed processing conditions when executing voice synthesis processing. In the example of this drawing, an example is shown in which voice quality setting, speed setting, and unknown word display setting of voice data are possible. As the processing conditions that can be set in the voice synthesis processing condition setting column 83, various settings such as setting the encoding speed (16 kbps, 32 kbps, etc.) of the generated voice data and the field of the word dictionary can be considered. .
[0016]
FIG. 4 shows an example of a voice data correction request screen on the client terminal 40. The voice data correction request screen 84 displays a voice data reproduction button 85 for instructing reproduction of voice data using the speaker 52, a correction input box 86 for prompting a character input as a correction instruction for text data, and an unknown word list. An unknown word list column 87, a word information input column 88 for prompting the input of the word information as a correction instruction for the unknown word, and an OK message button 89 for approving or giving a correction instruction to the voice data are included. The word information input field 88 further includes a reading input field 881 for prompting the user to input the reading of the word selected on the unknown word list 87, that is, “reading”, and a part of speech specification field 882 for specifying the part of speech. And an accent designation field 883 for prompting designation of the accent information in accordance with the reading specified in the reading input field. Note that a button for distributing approval and a button for instructing correction may be provided separately.
[0017]
FIG. 5 shows a processing procedure of the voice data synthesis server 20 and the client terminal 40. The sequence shown in this drawing will be described with reference to the components shown in FIGS. 1 and 2 as appropriate.
First, the voice data synthesizing server 20 transmits a voice data distribution request screen to the client terminal 40 by the HTTP / CGI unit 24 (step S11). This is transmitted when the client terminal 40 specifies the address of the voice data synthesis server 20 and accesses the voice data distribution request screen. In response to this transmission, the client terminal 40 receives and displays the audio data distribution request screen via the Web browser unit 44 (step S12). Next, text data input is received from the operator of the client terminal 40 and transmitted to the voice data synthesis server (step S13). In response to this transmission, the audio data distribution server 20 receives the text data via the HTTP / CGI unit 24 (Step S14).
[0018]
Next, the speech data synthesis server 20 executes the speech synthesis processing on the received text data by the speech synthesis processing unit 25 (step S15). In this synthesizing process, the synthesizing process may be performed according to the detailed voice synthesizing process conditions specified on the voice data distribution request screen. The speech synthesis processing unit 25 creates the speech data 63 and the unknown word list 64 as a result of the synthesis processing. Next, the voice data synthesizing server 20 transmits a voice data correction request screen including the obtained voice data 63 and the unknown word list 64 to the client terminal 40 by the HTTP / CGI unit 24 (step S16). In response to this transmission, the client terminal 40 receives and displays the voice data correction request screen via the Web browser unit 44 (step S17). In response to the display of the voice data correction request screen, the operator of the client terminal 40 appropriately determines whether or not the synthesized voice data is appropriate, and edits the text data as necessary.
[0019]
To explain an example of this editing, the text data "I can't hear the word I'm talking about here" is contextually a break "I can't hear the word I'm talking about here." If a sound that can be heard at this side is not heard, the operator of the requester terminal 40 adds a period to the original text data, and says, "I can hear the words spoken at this side. There is no correction instruction ", and a request for speech synthesis processing can be made again to generate an appropriate speech. Also, while the text data "He has Yamato soul" is appropriate in the context of reading "He has Yamato Damashii", if the sound that sounds "He has Daiwa Masai" is generated, the operator Can change the "Yamato soul" part of the original text data to "Yamatodamasashi" and Kana, and issue a correction instruction to generate an appropriate voice. In this case, even if the operator gives a correction instruction by inputting the word information of “Yamato soul” from the word information input field 88 shown in FIG. 4, an appropriate voice can be generated. The word information input here is used together with the word dictionary data 251 when the speech synthesis processing unit 25 performs speech synthesis processing again.
[0020]
After the above editing operation, the client terminal 40 transmits a voice data correction instruction (correction request) or an OK message to the voice data synthesis server via the web browser unit 44 (step S18). An OK message without a correction instruction means that the operator of the requester terminal 40 approves the actual distribution of the audio data without requiring a correction instruction.
[0021]
In response to this transmission, the audio data synthesizing server 20 receives a correction instruction (correction request) or an OK message through the HTTP / CGI unit 24 (step S19). Next, it is determined whether the instruction is a correction instruction / OK (step S20). If there is a correction instruction, the process returns to the speech synthesis processing in step S15. On the other hand, if it is OK, the audio data synthesizing server 20 causes the audio data distribution server 10 to store the audio data via the audio data accumulation requesting unit 26 (step S21). The voice data storage request unit 26 requests the voice data storage unit 15 of the voice data distribution server 10 to store voice data via the Internet 30, and the voice data storage unit 15 stores the request in the voice data DB 60. Thereby, the stored voice data can be delivered to the user terminal 100 or the client terminal 40 via the Internet 30 by the HTTP unit 14 of the voice data delivery server 10.
[0022]
Next, the voice data synthesizing server 20 notifies the client terminal of a registration completion message by the HTTP / CGI unit 24 (step S22). In response to this transmission, the client terminal 40 receives the registration completion message via the Web browser unit 44 (Step S23). This registration completion message is displayed on the client terminal 40 to the operator.
[0023]
As described above, in the first embodiment, since the voice data distribution system is provided with the voice synthesis server, the requester of the voice data distribution does not need to provide a voice synthesis function in the client terminal. It is possible to make appropriate corrections and additions to the synthesized voice data.
Although the distribution by the HTTP server has been described as the distribution form of the audio data, other distribution forms such as a file transfer v by FTP (File Transfer Protocol) and a distribution form by UDP (User Datagram Protocol) are not limited to this form. But it's fine.
<Second embodiment>
FIG. 7 is a second embodiment of the present invention, and shows another configuration of the audio data synthesizing server 20. Here, the hardware configuration of the audio data synthesizing server 20 is the same as that of the first embodiment, and the software configuration has some functions added. A description will be given only of a part where the function is added. Referring to FIG. 7, an identifier determining unit 27 and a voice / sound data information calculating unit 28 are provided.
[0024]
The identifier determining unit 27 automatically determines a unique identifier that can be accessed on the Internet 30 as audio data files of audio data synthesized and distributed based on a distribution approval OK message from the client terminal. Having. As a format of the identifier, a request source IP address (Internet Protocol; Source-address; IETF: RFC791) of an IP header can be used as information for specifying a URI (Uniform Resource Identifier; IETF: RFC2396).
[0025]
As an example of an automatic identifier generation method of the identifier determination unit 27, there is a method including a counter memory. In this example, a counter memory is provided inside the identifier determination unit 27. The identifier determination unit 27 reads the value from the counter memory immediately after starting the automatic generation. Then, an identifier including the character string of this value is generated, and the value of the counter memory is incremented. The value in the counter memory is stored until the next identifier generation. Specifically, when the value read from the counter memory is 123, for example, http: // aaa. bbb. cccc. dddd / 123. By generating a wav and an identifier and immediately incrementing the counter, the value 124 is prepared for the generation of the next identifier. As another example of the automatic generation method, there is a method using time information. In this example, a clock that provides time information is provided inside the identifier determination unit 27. Immediately after the automatic generation is started, the identifier determining unit 27 acquires time information from the clock and generates an identifier that includes a character string of this value. Specifically, when the time information obtained from the clock is 10: 29: 15: 33 ms on December 3, 2001, for example, the following is performed.
[0026]
http: // aaaa. bbbb. cccc. dddd / 2001120310291533. wav
And an identifier. The identifier generation function of the identifier determination unit 27 may be realized so as to automatically generate an identifier from the content of the target text data. As a method in this case, it is conceivable to preliminarily define the format of the text data and automatically extract a character example for which an identifier is to be expected from a character example in a predetermined field. This allows the requester to specify a desired identifier name by himself. This allows the requester to manage the identification of the voice data on the Internet by the identifier name specified by himself, thereby providing an easy-to-use system. The audio data storage requesting unit 26 of the audio data synthesizing server 20 assigns the identifier determined by the identifier determining unit 27 to the audio data 63 and requests the audio data distribution server 10 to store.
[0027]
The voice data information calculation unit 28 calculates voice data information based on the voice data synthesized by the voice synthesis processing unit 25. Here, the audio data information is the data size of the audio data and the reproduction time length of the audio data. The audio data information calculation unit 28 calculates the reproduction time of the audio data from the size of the input audio data and a predetermined audio data format. Specifically, assuming that the audio data size is N bytes, the format is a sampling frequency of 8 kHz, and the quantization is 16 bits, the reproduction time is calculated as N / (16) [ms].
[0028]
The audio data synthesizing server 20 transmits the reproduction time calculated by the audio data information calculator 28 to the client terminal 40 together with the identifier output from the identifier determiner 27 using the HTTP / CGI unit 24.
FIG. 7 shows an example of a voice data registration confirmation screen displayed on the client terminal 40. The audio data registration confirmation screen 90 includes an identifier display column 91 and a reproduction time display column 92. The identifier display column 91 displays the identifier determined by the identifier determination unit 27 of the audio data synthesis server 20. The reproduction time display column 92 displays the predicted reproduction time of the audio data calculated by the audio data information calculation unit 28 of the audio data synthesis server 20.
[0029]
In the second embodiment described above, the operator of the client terminal can recognize the identifier of the voice data as a result of the operation of the processing procedure similar to that of the first embodiment. For example, a Web page administrator updates this identifier by embedding it as a hyperlink in a Web page managed by the Web page manager, and registers the identifier in a Web server. When the user of the Web page displays the updated Web page on a browser or the like and wants to output the sound, the user can obtain the sound output by using the added hyperlink. Further, in the second embodiment, it is possible to recognize the reproduction time information of the audio data. As a result, it is possible to know the listening time and the amount of communication data when voice distribution is performed, and it is possible to assume a specific use image of the voice data user, and more sophisticated voice data distribution can be achieved. It becomes possible.
[0030]
In the first and second embodiments, a single client terminal or user terminal has been described for ease of explanation. A terminal or a user terminal is assumed. Therefore, the audio data that can be handled by the audio data distribution system is not limited to the illustrated number, but can include a large number of audio data.
[0031]
Further, the audio data distribution device in the first and second embodiments is constituted by two servers, namely, an audio data distribution server and an audio data synthesizing server. This is because the system can be easily realized by adding a new voice configuration server to an operation mode in which only a server exists. However, a configuration in which these two servers are integrated into a server device as a single piece of hardware is naturally possible.
[0032]
【The invention's effect】
As described above, according to the audio data distribution device and the distributor terminal according to the present invention, it is possible to confirm the reproduction of audio data to be distributed and make appropriate corrections, thereby avoiding utterance caused by erroneous reading. it can. As a result, a reliable audio data distribution device is provided.
[Brief description of the drawings]
FIG. 1 is a first embodiment of the present invention, and is a block diagram illustrating an overall configuration of an audio data distribution system.
FIG. 2 is a block diagram showing a configuration of a client terminal shown in FIG.
FIG. 3 is a diagram showing an example of a voice data distribution request screen displayed on a client terminal 40;
FIG. 4 is a diagram showing an example of a voice data correction request screen displayed on the client terminal 40.
FIG. 5 is a sequence diagram showing a processing procedure of the voice data synthesis server 20 and the client terminal 40.
FIG. 6 is a block diagram showing another configuration of the audio data synthesizing server 20 according to a second embodiment of the present invention.
FIG. 7 is a diagram showing an example of a voice data registration confirmation screen displayed on the client terminal 40.
[Explanation of symbols]
10 Audio data distribution server
12, 22, 41 control program
13,23,42 TCP / IP section
14 HTTP section
15 Voice data storage
20 Voice data synthesis server
24 HTTP / CGI section
25 Voice synthesis processing unit
26 Voice Data Storage Request Section
27 Identifier determination unit
28 Voice Data Information Calculation Unit
30 Internet
40 Requester terminal
43 I / O control unit
44 Web browser section
45 audio data playback unit
46 Voice input section
51 Display
52 speakers
53 microphone
54 keyboard
60 Voice Data Database
63 audio data
64 unknown word list
80 Voice data distribution request screen
84 Voice data correction request screen
90 Voice data registration confirmation screen
100 user terminals

Claims (7)

An audio data distribution device that distributes audio data via a data network,
Voice data storage means for receiving at least one set of text data from at least one client terminal, converting the text data into a set of voice data by voice synthesis, and storing the data for distribution;
Voice data correction updating means for correcting and updating the stored series of voice data in response to a correction request from the client terminal;
An audio data distribution device comprising: The voice data correction updating unit corrects and updates the series of voice data based on a correction request including correction information for text data corresponding to the series of voice data and / or word information for an unknown word. The audio data distribution device according to claim 1, wherein: 3. The voice data according to claim 1, further comprising: identifier determination means for determining an identifier of the series of voice data; and identifier transmission means for transmitting the determined identifier to a corresponding client terminal. Delivery device. 4. The audio data distribution device according to claim 3, wherein the identifier determination unit determines the identifier based on a counter updated for each series of audio data, a current time, or information extracted from the text data. 2. The audio data distribution device according to claim 1, further comprising: means for calculating an estimated reproduction time of the series of audio data and transmitting the estimated reproduction time to a corresponding client terminal. A requester terminal for requesting an audio data distribution device that distributes audio data via a data network to distribute at least one series of audio data of the plurality of audio data,
Text data transmitting means for transmitting the text data to be converted to the series of voice data to the voice data distribution server;
Correction information transmitting means for transmitting a correction request for correcting and updating the series of voice data to the voice data distribution server;
A distribution requester terminal comprising: 7. The distribution requester terminal according to claim 6, wherein the correction request includes correction information for text data corresponding to the series of voice data and / or word information for unknown words.

Images