JP2013057887A - Voice output device, voice output method, and voice output program - Google Patents

Abstract

【Task】
In an audio output device that outputs multilingual audio, waste of information provision caused by not understanding the language used by the listener group is reduced.
[Solution]
An audio output content input unit for inputting audio output content; an output audio generation unit for generating output audio composed of a plurality of languages based on the audio output content; an environmental audio input unit for inputting audio of the surrounding environment; A language composition ratio calculation unit that recognizes the language of the input speech and calculates a language composition ratio, and an audio output configuration that changes the structure of the output sound according to the language composition ratio calculated by the language composition ratio calculation unit A changing unit; and a sound output unit that outputs the changed sound.
[Selection] Figure 1

Description

  The present invention relates to an audio output device such as an automatic broadcast system, and more particularly to an audio output device that outputs multilingual audio.

As a background technology of this technology, there is an automatic broadcasting system in public facilities.
Upon receiving a broadcast command, the automatic broadcast system can output a broadcast by combining pre-recorded guidance voices. The broadcast command includes designation of broadcast contents, broadcast time, or broadcast location, and realizes a variety of broadcasts according to the situation. In some systems, a speech synthesizer that converts text into speech is used, and it is possible to generate not only fixed phrases but also various contents of speech simply by giving text as a broadcast command.

  In recent years, the importance of internationalization in public facilities has expanded. Broadcasting at international event venues, airports, railway stations, museums, etc. requires voice guidance in multiple languages. The automatic broadcasting system realizes multilingual voice guidance by recording multilingual guidance voices in advance or using a multilingual voice synthesizer.

  However, as one problem in such an automatic broadcast system, there is a problem that the time required for voice guidance becomes long. If another language is added to a system that has conventionally realized broadcasting in a single language, it is inevitable that the broadcasting will be long. In addition, since broadcasting in multiple languages broadcasts all languages continuously regardless of the language used by the listener, most of the broadcasting portions provide unnecessary information.

  In order to reduce the waste of providing information due to not understanding the language of the listener, a system that provides a mobile terminal device and earphones and provides audio guidance only in a specific language is a museum. Etc. However, in such a system, it is necessary to have each customer attach a mobile terminal and an earphone, which is troublesome. In addition, there is a weak point that it lacks the function to inform the group.

  As a method for understanding the language used by the listener, there is a method using language recognition. For example, in Patent Document 1, in order to determine a language used in an in-vehicle navigation device, “for each word included in the input speech, a corresponding language is investigated, and a language that occupies a large number in the survey result is the language used. Judgment as ".

JP 2001-296882 A

P. A. Torres-Carrasquillo et al. , “Approaches to language identification using Gaussian mixture models and shifted delta cepstral features.” Proc. ICSLP 2002, Sept. 2002, pp. 89-92.

  However, in the technique disclosed in Patent Document 1, although only one language can be determined for the observed language, it is not possible to understand the multiple languages of the listener group, and the multilingual voice guidance is controlled. I can't do it.

  An object of the present invention is to provide an audio output device that outputs multilingual audio corresponding to a plurality of languages used by a group of listeners and reduces waste of information provision.

  In order to solve the above problems, for example, the configuration described in the claims is adopted.

  If an example of the audio output device of the present invention is given, an audio output content input unit that inputs audio output content, an output audio generation unit that generates output audio composed of a plurality of languages based on the audio output content, and surroundings Depending on the language composition ratio calculated by the environment composition input section that inputs the environment sound, the language composition ratio calculation section that recognizes the language of the input sound and calculates the composition ratio of the language, and the language composition ratio calculation section The audio output device includes an audio output configuration change unit that changes the configuration of the output audio and an audio output unit that outputs the changed audio.

  The audio output device of the present invention further includes an audio output language designating unit for designating an audio output language, and the output audio generating unit generates output audio based on the audio output content and the audio output language designation. In addition, the language composition ratio calculation unit may recognize a speech language based on the speech output language designation.

  The audio output device of the present invention further includes a position information acquisition unit that acquires position information of the audio output device, and an audio output language storage unit that stores an audio output language in association with the position information. The output language designation unit may designate an audio output language acquired from the audio output broadcast language storage unit according to the position information.

  The voice output device of the present invention further includes a configuration change method storage unit that stores a method of changing a configuration of voice to be output in association with a language composition ratio, and the voice output configuration change unit includes the language configuration According to the language composition ratio calculated by the ratio calculation unit, the configuration of the output voice may be changed based on the change method acquired from the configuration change method storage unit.

  The audio output device of the present invention further includes an audio output configuration generation unit that generates an audio output configuration based on the audio output language and the output audio, and the configuration change method storage unit includes the audio output configuration and the language. A method of changing a configuration of audio to be output in association with a composition ratio is stored, and the audio output configuration change unit is acquired from the configuration change method storage unit according to the audio output configuration and the language configuration ratio. The configuration of the output voice may be changed based on the changed method.

  In the speech output device of the present invention, the speech output content input unit inputs text of speech output text, and the output speech generation unit generates output speech by speech synthesis processing for converting the text into speech, The speech output configuration changing unit may perform speech synthesis processing based on edited text obtained by rearranging text, specifying speech speed for a part of text, or deleting part of text.

  If an example of the audio output method of the present invention is given, an audio output content input step for inputting audio output content, an output audio generation step for generating output audio consisting of a plurality of languages based on the audio output content, and surroundings In accordance with the environment composition input step for inputting environment speech, the language composition ratio calculation step for recognizing the language of the input speech and calculating the language composition ratio, and the language composition ratio calculated in the language composition ratio calculation step Thus, there is provided an audio output method including an audio output configuration change step for changing an audio configuration to be output, and an audio output step for outputting the changed audio.

  The audio output method of the present invention further includes an audio output language designation step for designating an audio output language, and the output audio generation step generates an output audio based on the audio output content and the audio output language designation. The language composition ratio calculating step may recognize a speech language based on the speech output language designation.

  The audio output method of the present invention further includes an audio output configuration generation step for generating an audio output configuration based on the audio output language and the output audio, and the audio output configuration change step includes the audio output configuration In accordance with the language composition ratio, the configuration of the output voice is changed based on the storage contents of the configuration change method storage unit storing the voice output configuration and the method of changing the configuration of the voice to be output in association with the language configuration ratio. Things can be used.

  An example of the audio output program of the present invention is an audio output program for causing a computer to output audio, an audio output content input step for inputting audio output content, and an audio output language for specifying an audio output language. A designation step, an output voice generation step for generating an output voice composed of a plurality of languages based on the voice output contents and the voice output language designation, an environmental voice input step for inputting a voice of the surrounding environment, and the voice output language designation The language composition ratio calculating step for recognizing the language of the input speech based on the language and calculating the language composition ratio, and changing the configuration of the output speech according to the language composition ratio calculated in the language composition ratio calculation step An audio output program for executing an audio output configuration change step and an audio output step for outputting the changed audio It is a non.

  The audio output program of the present invention further includes an audio output configuration generation step for generating an audio output configuration based on the audio output language and the output audio, and the audio output configuration change step includes the audio output configuration and the language. The configuration of the output voice is changed based on the storage contents of the configuration change method storage unit that stores the method of changing the configuration of the voice to be output in association with the voice output configuration and the language configuration ratio according to the configuration ratio. Good

  According to the present invention, multilingual voice guidance can be controlled based on the language composition ratio of a listener group, and information provision waste caused by not understanding the language used by the listener group can be reduced. be able to.

It is a figure explaining the whole broadcast output system composition of Example 1. It is a hardware block diagram of the broadcast output system of Example 1. 6 is a flowchart illustrating an operation of environmental sound accumulation processing according to the first embodiment. 3 is a flowchart illustrating an operation of broadcast output processing in the first embodiment. It is a figure which shows an example of a structure change method storage table. It is a figure explaining the whole structure of the broadcast output system of Example 2. FIG. It is a hardware block diagram of the broadcast output system of Example 2. 10 is a flowchart illustrating an operation of broadcast output processing in the second embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  The present embodiment is an example in which the audio output device is applied to a broadcast output system in which the language composition ratio of the listener group is extracted from the environmental sound and the broadcast language composition is changed based on the language composition ratio.

FIG. 1 is a diagram for explaining the overall configuration of the broadcast output system of the present embodiment.
The broadcast output system 1 includes a broadcast generation device 2 and a broadcast output device 3. The broadcast generation device 2 includes a broadcast content input unit 4, a broadcast language designation unit 5, a broadcast audio generation unit 6, a broadcast configuration generation unit 7, and a transmission unit 15. The broadcast output device 3 includes an environmental audio input unit 8, an environmental audio storage unit 9, a language recognition unit 10, a language composition ratio calculation unit 11, a configuration change method storage unit 12, a broadcast configuration change unit 13, a broadcast audio output unit 14, and a reception The unit 16 is provided.

  The broadcast content input unit 4 receives input of information indicating broadcast content. The broadcast language designation unit 5 determines the broadcast language according to the broadcast content. The broadcast audio generation unit 6 generates broadcast audio based on the designation of the broadcast content / broadcast language. The broadcast configuration generation unit 7 extracts the language configuration included in the broadcast audio. The transmission unit 15 transmits the broadcast language designation, the broadcast audio, and the language configuration included in the broadcast audio to the broadcast output device 3.

  The environmental sound input unit 8 receives an input of environmental sound, and the environmental sound storage unit 9 stores the environmental sound for a predetermined time.

  The receiving unit 16 receives a broadcast language designation, broadcast audio, and a language configuration included in the broadcast audio transmitted from the broadcast generation device 2. The language recognition unit 10 performs language recognition of the environmental sound stored in the environmental sound storage unit 9 based on the designation of the broadcast language. The language composition ratio calculation unit 11 calculates a language composition ratio based on a time series of language recognition results for the accumulated environmental speech. The broadcast configuration change unit 13 acquires the configuration change method stored in the configuration change method storage unit 12 using the language composition ratio of the environmental sound and the broadcast configuration as inputs, and changes the broadcast audio based on the configuration change method. The broadcast audio output unit 14 outputs the changed broadcast audio.

  FIG. 2 is an example of a hardware configuration diagram of the broadcast output system of the present embodiment. The broadcast generation device 2 and the broadcast output device 3 are connected through the network device 99. The broadcast generation device 2 includes a keyboard 101, a mouse 102, a display 103, and a communication device 104. A broadcast content input process 105, a broadcast language designation process 106, a broadcast audio generation process 107, a broadcast configuration generation process 108, a sound source data 109, and a transmission process 110 are provided in the memory 111, and the CPU 112 executes the program. For example, it can be configured using a personal computer.

  The broadcast output device 3 includes a microphone 121, a speaker 122, and a communication device 123. Environmental audio input processing 124, environmental audio storage processing 125, language recognition processing 126, environmental audio storage data 127, language composition ratio calculation processing 128, configuration change method storage table 129, broadcast configuration change processing 130, broadcast audio output processing 131, reception The processing 132 is provided in the memory 133, and the CPU 134 executes the program. For example, it can be configured using a personal computer.

  The broadcast output system 1 can be realized by the CPU 112 loading the memory 111 and the CPU 134 executing a predetermined program loaded on the memory 133. Although this predetermined program is not shown in the figure, it is input from a storage medium storing the program via a reading device, or from a network via a communication device, and is directly loaded on a memory, or Once stored in the external storage device, it may be loaded onto the memory.

  The invention of the program in the present invention is a program that is incorporated in a computer as described above and causes the computer to operate as an audio output device. The audio output apparatus shown in the block diagram of FIG. 2 is configured by incorporating the program of the present invention into a computer.

  Next, the operation of the broadcast output system 1 according to the present embodiment configured as described above will be described. The operation of this system is divided into environmental sound accumulation processing and broadcast sound output processing.

The operation related to the environmental sound accumulation processing will be described with reference to FIG.
FIG. 3 is a flowchart illustrating the operation of the environmental sound accumulation process according to the first embodiment.

  First, the environmental sound input unit 8 acquires environmental sound from the microphone 121 by executing the environmental sound input process 124 (step S1). Here, the sound input to the microphone 121 includes the sound of the listener group. For this purpose, the microphone 121 may be placed not only near the speaker 122 but also in another place where a similar group of listeners exists. For example, when considering broadcasting at the boarding gate in airport announcements, it is expected that it will be difficult to determine the language used by the listener group due to various noises mixed near the boarding gate. If it is installed in the lounge, it is possible to acquire sound in a relatively quiet environment.

  Next, the environmental sound storage unit 9 stores the acquired environmental sound in the environmental sound storage data 127 for a predetermined time by executing the environmental sound storage processing 125 (step S2). Here, the time for which the sound is accumulated is a predetermined time set in the broadcast output device 3 in advance. For example, the time is 30 seconds, 1 hour, or 1 day. For example, if it is 30 seconds, processing corresponding to a listener that has been relatively close to the broadcast output device 3 is performed, and this can be applied to a local broadcasting system for a conference room in which the listener is replaced in a short time. Moreover, if it is 1 hour, the process according to a group of several listeners will be performed in a longer time. It can be applied to announcements of event venues that change customers on a daily basis. Alternatively, when a long time such as one day is set, automatic language configuration optimization can be achieved for a system such as a railway station where a listener is fixed to some extent.

  In order to change the accumulation time, it has an accumulation time designation unit that designates the accumulation time of the environmental sound, and the environmental sound accumulation unit changes the predetermined time for accumulating the environmental sound based on the designation of the accumulation time designation unit What is necessary is just to comprise so.

The operation related to the output processing of the broadcast audio will be described with reference to FIG.
FIG. 4 is a flowchart illustrating the operation of broadcast sound output processing in the first embodiment.

  First, the broadcast content input unit 4 determines the broadcast content by executing the broadcast content input process 105 (step S101). For example, in a local broadcasting system at a railway station, the broadcast content is determined using a train number, arrival time, etc. included in the operation management information received from the communication device 104. Also, in the case of a local broadcasting system that guides event information, guidance candidates are displayed on the display 103 and selected by the keyboard 101 or the mouse 102 to determine the broadcast content.

  Next, the broadcast language designation unit 5 designates the broadcast language by executing the broadcast language designation process 106 (step S102). The broadcast language designation is received from the communication device 104 by the same means as in step S101, or the language candidate is displayed on the display 103 and designated by selecting with the keyboard 101 or the mouse 102.

  Next, the broadcast audio generation unit 6 generates broadcast audio using the sound source data 109 based on the input broadcast content and broadcast language designation by executing the broadcast audio generation processing 107 (step S103). Here, broadcast sound can be generated by a general recording editing method. For example, “Broadcast type: Arrival, Train number: Hikari 23, Departure time: 10:30, Destination: Tokyo, Arrival number: 1” is given as the broadcast content of the on-site broadcast at the railway station, and the broadcast language is “English”・ If Japanese is specified, “Shortly, the special rapid service train,“ Hikari ”23, departing at 10:30, bound for Tokyo, will be arriving at Track 1. For your safety, please step back from the "Soon, there will be a train for 10:30, Hikari 23, and Tokyo bound for Line 1. Please wait inside the yellow line."

  Next, by executing the broadcast configuration generation process 108, the broadcast configuration generation unit 7 measures the duration of each language included in the broadcast audio, and outputs it along with the language order (step S104). For example, in the above example, the information is generated in the order of English and Japanese, and the configuration information is output as, for example, “English: 20 seconds → Japanese: 18 seconds”.

  Next, the transmission unit 15 executes the transmission process 110 and transmits the broadcast language designation, the broadcast sound, and the broadcast configuration to the broadcast output device 3 via the network device 99 using the communication device 104. (Step S105).

  Next, the reception unit 16 executes the reception process 132 and receives the broadcast language designation, the broadcast sound, and the broadcast configuration from the broadcast generation device 2 via the network device 99 using the communication device 123. (Step S106).

Next, the language recognizing unit 10 acquires the environmental sound from the environmental sound accumulation data 127 by executing the language recognition process 126, and performs language recognition based on the designation of the broadcast language (step S107). For the language recognition in this step, a known language recognition technique can be used, but the language recognition result is preferably a time series determined for each unit time frame. For example, as a method for outputting a language recognition result for each unit time frame, a method based on a mixed Gaussian distribution model of Torres-Carrasquillo et al. Can be used. (Non-Patent Document 1)
Specifically, a cepstrum feature amount is calculated for each unit time frame, and inter-frame difference information is added to obtain a frame feature amount. Next, the likelihood for the mixed Gaussian model learned in advance for each language is obtained based on the designation of the broadcast language, and the language having the maximum likelihood is used as the determination result.

  Next, the language composition ratio calculation unit 11 calculates the language composition ratio based on the time series of the language recognition results (step S108). Here, if the above mixed Gaussian distribution model is used, language determination results are obtained for each short-time frame. By summing up the results, the language determination frequency in the accumulated environmental speech is used as the language determination frequency. The composition ratio is obtained. For example, “English: 0%, Japanese: 100%” is output. In addition to using a mixed Gaussian model, language recognition methods include a method using phoneme recognition and a language model, and a method using large vocabulary continuous speech recognition. In such a language recognition method, a language determination result is given as a recognition result in units of phonemes, words, and phrases. In that case, in order to be able to aggregate the determination results for each unit time, record the time length of the target phoneme / word / phrase together with the language determination result. You just have to ask for it.

  Next, the broadcast configuration change unit 13 performs the configuration change method storage table based on the language configuration ratio obtained by the language configuration ratio calculation processing 128 by the execution of the broadcast configuration change processing 130 and the broadcast configuration received by the reception unit 16. The corresponding configuration change method is acquired from 129, and the configuration of the broadcast audio is changed based on the configuration change method (step S109).

FIG. 5 is an explanatory diagram showing an example of the configuration change method storage table 129. In this example, in the first line, if the broadcast composition 201 is “English: 70%, Japanese: 30%” and the environmental language composition ratio 202 is “English: 0%, Japanese: 0%”, As the change method 203, it is specified to reduce the speaking speed in both English and Japanese. On the second line, if the broadcast composition 201 is “English: 70%, Japanese: 30%” and the environmental language composition ratio 202 is “English: 0%, Japanese: 100%”, the change method As 203, Japanese is changed to be broadcast first, and English is specified to be shortened by summary. If there is a match on the second line, the broadcast content “Shortly, the special rapid service train,“ Hikari ”23, departing at 10:30, bound for Tokyo, will be arriving at” Track 1. For your safety, please step back from the edge of the platform Soon we will be on the 1st line with a 10:30 train, Hikari 23, and a train bound for Tokyo. "For example," Soon, 10:30, Hikari 23, train to Tokyo will come to line 1. Because it is dangerous, please wait inside the yellow line. Shortly, the train will be arriving at Track. " 1 ”is changed.
Here, as a change method, when changing the speech speed, a known speech speed conversion method can also be used. In addition, as a method for summarizing, a method may be used in which broadcast sections that can be deleted are designated in advance as broadcast configuration information.

  Finally, the broadcast audio output unit 14 outputs the broadcast audio changed by the broadcast configuration change process 130 from the speaker 122 by executing the broadcast audio output process 131.

  As described above, the broadcast output system of the present embodiment can extract the language composition ratio of the listener group from the environmental sound and change the broadcast composition based on the language composition ratio.

  In this embodiment, the environmental sound acquired by the environmental sound input unit 8 is accumulated in the environmental sound storage unit 9 for a predetermined time, and the accumulated environmental sound is recognized as a language. The environmental sound acquired in step 1 may be recognized in a language for a predetermined time.

  Further, although the language recognition unit 10 performs language recognition based on the language designated by the broadcast language designation unit 5, language recognition may be performed without designating the language.

  In the present embodiment, the broadcast generation device 2 and the broadcast output device 3 are separate devices and are connected by a network device. However, the broadcast generation device 2 and the broadcast output device 3 may be a single device. .

  In the first embodiment, an example in which language recognition is performed based on the language configuration of broadcast audio has been described. However, the language distribution can be inferred from the position information. Therefore, by performing language recognition based on the position information, a highly accurate language composition ratio can be obtained. Further, in the first embodiment, an example in which broadcast sound is generated by a recording and editing method has been described. However, it is possible to edit broadcast contents more flexibly by generating broadcast sound by a text-to-speech synthesis method. In the first embodiment, the broadcast content is determined by input to the broadcast generation device. However, a configuration in which the broadcast content (text) is generated on the broadcast output device side is also possible. In this embodiment, an example applied to a broadcast output system using such position information and text-to-speech synthesis method will be described.

FIG. 6 is a diagram for explaining the overall configuration of the broadcast output system of the present embodiment.
The broadcast output system 301 includes a broadcast generation device 302 and a broadcast output device 303. The broadcast generation apparatus 302 includes a broadcast content reception unit 304, a language composition ratio reception unit 305, a configuration change method storage unit 306, a text generation unit 307, a broadcast configuration extraction unit 308, a broadcast configuration editing unit 309, a voice synthesis unit 310, and a communication server. 311 is provided. The broadcast output device 303 includes an environmental audio input unit 320, an environmental audio storage unit 321, a position information acquisition unit 322, a language candidate setting unit 323, a language recognition unit 324, a language composition ratio calculation unit 325, a broadcast information storage unit 326, and broadcast contents. A generation unit 327, a communication client unit 328, and a broadcast audio output unit 329 are provided.

The environmental sound input unit 320 accepts input of environmental sound, and the environmental sound storage unit 321 stores the environmental sound for a predetermined time.
The position information acquisition unit 322 acquires the position information of the broadcast output device 303 and sets a language candidate corresponding to the position information in the language candidate setting unit 323.
The language recognition unit 324 performs language recognition based on the designation of language candidates in the language candidate setting unit 323. The language composition ratio calculation unit 325 generates a language composition ratio based on a time series of language recognition results for the accumulated environmental speech.
The broadcast information storage unit 326 is a storage area for managing broadcast time and broadcast content. The broadcast content generation unit 327 manages the time and acquires the broadcast content according to the broadcast time from the broadcast information storage unit.
The communication client unit 328 transmits the broadcast content and the language composition ratio to the broadcast generation apparatus 302 and receives broadcast audio from the broadcast generation apparatus 302.
Broadcast audio output unit 329 outputs the received broadcast audio.

The communication server 311 receives the broadcast content and the language composition ratio from the broadcast output device 303 and transmits the broadcast sound generated as a response to the broadcast output device 303.
The broadcast content receiving unit 304 acquires broadcast content from the message received by the communication server 311. The language distribution receiving unit 305 acquires the language composition ratio from the message received by the communication server 311.
The text generation unit 307 generates text for multilingual voice guidance according to the received broadcast content.
The broadcast configuration extraction unit 308 extracts the language configuration included in the broadcast audio when the generated text is synthesized.
The broadcast configuration editing unit 309 receives the received language composition ratio and the extracted broadcast configuration, acquires the configuration change method stored in the configuration change method storage unit 306, and edits the text based on the configuration change method. .
The voice synthesizer 310 performs voice synthesis based on the edited text, and generates broadcast voice for the communication server 311 to transmit as a response.

  FIG. 7 is an example of a hardware configuration diagram of the broadcast output system of the present embodiment. The broadcast generation device 302 and the broadcast output device 303 are connected through a network device 499. The broadcast generation device 302 includes a communication device 413. A text generation process 404, a broadcast configuration extraction process 405, a broadcast configuration edit process 406, a speech synthesis process 407, synthesis data 408, a configuration change method storage table 409, and a communication process 410 are provided in the memory 411, and the CPU 412 executes the program. For example, it can be configured using a personal computer.

  The broadcast output device 303 includes a microphone 421, a speaker 422, a keyboard 423, a mouse 424, a display 425, a GPS receiver 426, and a communication device 440. Environmental sound input processing 427, environmental sound storage processing 428, position information acquisition processing 429, language candidate setting processing 430, language recognition processing 431, language composition ratio calculation processing 432, environmental sound storage data 433, broadcast content generation processing 434, communication processing 435, a broadcast information storage table 436, and a broadcast audio output process 437 are provided in the memory 438, and the CPU 439 executes the program. For example, it can be configured using a personal computer.

Next, the operation of the broadcast output system 301 according to this embodiment configured as described above will be described.
The operation of this system is divided into environmental sound accumulation processing and broadcast sound output processing. The environmental sound accumulation process is the same as in the first embodiment.

FIG. 8 is a flowchart showing the operation of the broadcast output process in the present embodiment.
First, the position information acquisition unit 322 acquires the positioning information obtained from the GPS receiver 426 as the position information of the broadcast output device 303 by executing the position information acquisition process 429 (step S201).
Next, the language candidate setting unit 323 sets language candidates predetermined according to the position information by executing the language candidate setting process 430 (step S202). The language candidates used here are, for example, in railway broadcasting, the type of broadcast language required by the country or region is determined at the time of design. Therefore, the correspondence between position information and language candidates is described in the language candidate setting process 430. You just have to.
Next, by executing the broadcast content generation process 434, the broadcast content generation unit 327 generates text indicating the broadcast content with reference to the broadcast information storage table 436 (step S203). Here, a character string such as “broadcast type: arrival, train number: Hikari No. 23, departure time: 10:30, destination: Tokyo, arrival number line: 1” is taken as an example of the broadcast content. Alternatively, in a local broadcasting system that guides event information, guide candidates are displayed on the display 425 and selected with the keyboard 101 or the mouse 102, or the text of the broadcast content is directly input with the keyboard 101. The broadcast content may be determined accordingly.
Next, the language recognizing unit 324 acquires the environment sound from the environment sound accumulated data 433 by executing the language recognition process 431, and performs language recognition based on the language candidate (step S204). The language recognition method in this step is the same as that in the first embodiment.
Next, the language composition ratio calculation unit 325 calculates a language composition ratio based on the time series of the language recognition results (step S205). The specific method of this step is the same as that in the first embodiment.
Next, the communication client unit 328 executes communication processing 435, transmits the composition ratio between the text of the broadcast content and the calculated language from the communication device 440 to the broadcast generation device 302 via the network device 499, and sends a response. Wait (step S206).

Next, the broadcast content receiving unit receives the broadcast content from the communication device 413 by executing the communication process 410 by the communication server 311 (step S207).
Next, the text generation unit 307 generates text for text-to-speech synthesis based on the received text of the broadcast content by executing the text generation processing 404 (step S208). Here, when the received text is “Broadcast type: Arrival, Train number: Hikari 23, Departure time: 10:30, Destination: Tokyo, Arrival number: 1”, “Shortly, the special rapid service train, “Hikari” 23, departing at 10:30, bound for Tokyo, will be arriving at Track 1. For your safety, please step back from the edge of the platform. A train to Tokyo is coming. Please wait inside the yellow line because it ’s dangerous. ” Such a process is realized by performing a process performed in a general recording editing system by a combination of texts instead of voices.
Next, when the generated text is synthesized by text-to-speech synthesis by the execution of the broadcast configuration extraction process 405, the broadcast configuration extraction unit 308 measures the duration of each language included in the broadcast audio, and determines the language order. Together with this (step S209).
Next, the language composition ratio receiving unit 305 receives the language composition ratio from the communication device 413 by executing the communication process 410 by the communication server unit 311 (step S210).

Next, the broadcast configuration editing unit 309 performs the configuration change method storage table 409 based on the language configuration ratio indicated by the received language distribution and the broadcast configuration extracted by the broadcast configuration extraction unit 308 by executing the broadcast configuration editing process 406. The corresponding configuration change method is acquired from the above, and the language configuration of the broadcast audio is changed based on the configuration change method (step S211). Here, the configuration change method storage table 409 can be as shown in FIG. 5 as in the first embodiment.
However, here, unlike the first embodiment, the changing method is performed by editing text. That is, when changing the speech speed, the speech speed is changed using a method of giving speed as a parameter for text-to-speech synthesis. Moreover, in order to summarize, what is necessary is just to designate the range of text that can be deleted in advance as broadcast configuration information. Alternatively, text summarization processing may be used for simplification. In this case, the recording editing method can only be deleted, but it is also possible to obtain an expression with a shorter utterance time by changing the vocabulary or the like.

Next, the speech synthesizer 310 synthesizes broadcast speech from the edited text using the synthesis data 408 by executing the speech synthesis processing 407 (step S212). As the speech synthesis technology here, a known speech synthesis technology can be used. However, in order to realize a change in speech speed for each language, a parameter that can be partially changed in speed should be used. desirable.
Next, the communication server 311 transmits the synthesized broadcast audio from the communication device 413 to the broadcast output device 303 via the network device 499 by executing the communication process 410 (step S213).

Next, the communication client unit 328 receives broadcast sound by executing the communication process 435 (step S214).
Finally, the broadcast audio output unit 329 outputs the broadcast audio from the speaker 422 by executing the broadcast audio output process 437 (step S215).

  As described above, the broadcast output system according to the present embodiment extracts the language composition ratio of the listener group from the environmental sound, and changes the broadcast language composition based on the language composition ratio, in addition to the location information. This makes it easy to extract the language composition ratio, makes it possible to change the broadcast language structure more flexibly by using the text-to-speech synthesis method, and allows the broadcast output device side to input broadcast contents. .

  According to each of the embodiments described above, the language composition ratio of the listener group is extracted from the environmental sound, and the language used by the listener group is understood by changing the structure of the voice output based on the language composition ratio. It is possible to reduce the waste of providing information due to the absence.

  In each of the above-described embodiments, various functions of the broadcast output system are realized by a program executed on the CPU. However, some or all of them are hardware using electronic components such as an integrated circuit, for example. May be realized.

  The present invention is not limited to the above-described embodiments, and includes various modifications. Although the present embodiment has been described assuming a broadcasting system such as a railway or a public facility, it can be applied to various devices and systems that perform audio output, such as an audio communication system such as a mobile phone and a remote conference system. it can.

DESCRIPTION OF SYMBOLS 1 Broadcast output system 2 Broadcast production | generation apparatus 3 Broadcast output apparatus 4 Broadcast content input part 5 Broadcast language designation | designated part 6 Broadcast audio | voice production | generation part 7 Broadcast structure production | generation part 8 Environmental sound input part 9 Environmental sound storage part 10 Language recognition part 11 Language composition ratio Calculation unit 12 Configuration change method storage unit 13 Broadcast configuration change unit 14 Broadcast audio output unit 15 Transmission unit 16 Reception unit 301 Broadcast output system 302 Broadcast generation device 303 Broadcast output device 304 Broadcast content reception unit 305 Language composition ratio reception unit 306 Configuration change Method storage unit 307 Text generation unit 308 Broadcast configuration extraction unit 309 Broadcast configuration editing unit 310 Speech synthesis unit 311 Communication server 320 Environmental speech input unit 321 Environmental speech storage unit 322 Location information acquisition unit 323 Language candidate setting unit 324 Language recognition unit 325 Language Composition ratio calculation unit 326 Broadcast information storage unit 327 Broadcast content generation unit 328 Communication class Ant section 329 audio output unit

Claims (14)

An audio output content input unit for inputting audio output content;
An output audio generation unit that generates output audio composed of a plurality of languages based on the audio output content;
An environmental sound input unit for inputting sound of the surrounding environment;
A language composition ratio calculation unit that recognizes the language of the input speech and calculates a composition ratio of the language;
A voice output configuration changing unit that changes a configuration of voice to be output in accordance with the language component ratio calculated by the language component ratio calculating unit;
An audio output unit for outputting the changed audio;
An audio output device. The audio output device according to claim 1, further comprising:
Has an audio output language specification part for specifying an audio output language,
The output sound generation unit generates output sound based on sound output content and sound output language designation,
The language composition ratio calculation unit recognizes a speech language based on the speech output language designation. The audio output device according to claim 2, further comprising:
A position information acquisition unit for acquiring position information of the audio output device;
An audio output language storage unit that stores the audio output language in association with the position information;
The audio output language designating unit designates an audio output language acquired from the audio output broadcast language storage unit according to position information. The audio output device according to claim 1, further comprising:
An environmental sound storage unit that stores sound of a predetermined time input by the environmental sound input unit;
The language composition ratio calculation unit recognizes a language of the accumulated speech and calculates a language composition ratio. The audio output device according to claim 4, further comprising:
It has an accumulation time designation part that designates the accumulation time of environmental sound,
The environmental sound accumulation unit changes a predetermined time for accumulating environmental sound based on the designation of the accumulation time designation unit. The audio output device according to claim 1, further comprising:
A configuration change method storage unit that stores a method of changing the configuration of audio to be output in association with the language composition ratio;
The voice output configuration change unit changes the configuration of the output voice based on the change method acquired from the configuration change method storage unit according to the language configuration ratio calculated by the language configuration ratio calculation unit. Voice output device. The audio output device according to claim 6, further comprising:
An audio output configuration generation unit that generates an audio output configuration based on the audio output language and the output audio;
The configuration change method storage unit stores a method for changing the configuration of audio to be output in association with the audio output configuration and the language composition ratio,
The voice output configuration changing unit changes a configuration of voice to be output based on a change method acquired from the configuration change method storage unit according to the voice output configuration and the language configuration ratio. Output device. In the audio output device according to any one of claims 1 to 7,
The voice output content input unit inputs text of a voice output sentence,
The output speech generation unit generates output speech by speech synthesis processing for converting text into speech,
The voice output configuration changing unit performs voice synthesis processing based on edited text obtained by rearranging text, specifying speech speed for a part of text, or deleting part of text. apparatus. An audio output content input step for inputting audio output content;
An output sound generation step for generating an output sound composed of a plurality of languages based on the sound output content;
An environmental voice input step for inputting the voice of the surrounding environment;
A language composition ratio calculating step of recognizing a language of the input speech and calculating a language composition ratio;
A voice output configuration change step of changing a configuration of voice to be output according to the language configuration ratio calculated in the language configuration ratio calculation step;
An audio output step for outputting the changed audio;
An audio output method comprising: The audio output method according to claim 9, further comprising:
An audio output language designation step for designating an audio output language;
The output sound generation step generates output sound based on sound output content and sound output language designation,
The language composition ratio calculating step recognizes a speech language based on the speech output language designation. The audio output method according to claim 10, further comprising:
An audio output configuration generation step for generating an audio output configuration based on the audio output language and the output audio;
The voice output configuration change step stores a configuration change method storage unit that stores a method of changing a voice configuration to be output in association with the voice output configuration and the language configuration ratio according to the voice output configuration and the language configuration ratio. An audio output method characterized by changing a configuration of audio to be output based on content. The voice output method according to any one of claims 9 to 11,
The voice output configuration change step includes at least one or more of a process for changing the language order of voice output, a process for converting the speech speed for each language, and a summarization process for deleting a part of the voice content for each language to shorten it. An audio output method characterized by the above. An audio output program for causing a computer to output audio,
An audio output content input step for inputting audio output content;
An audio output language specification step for specifying an audio output language;
An output sound generation step of generating output sound composed of a plurality of languages based on the sound output content and the sound output language designation;
An environmental voice input step for inputting the voice of the surrounding environment;
A language composition ratio calculating step of recognizing the language of the input speech based on the sound output language designation and calculating a language composition ratio;
A voice output configuration change step of changing a configuration of voice to be output according to the language configuration ratio calculated in the language configuration ratio calculation step;
An audio output step for outputting the changed audio;
Audio output program for running 14. An audio output program for causing a computer to output audio according to claim 13, further comprising:
An audio output configuration generation step for generating an audio output configuration based on the audio output language and the output audio;
The voice output configuration change step stores a configuration change method storage unit that stores a method of changing a voice configuration to be output in association with the voice output configuration and the language configuration ratio according to the voice output configuration and the language configuration ratio. An audio output program for changing the configuration of audio to be output based on the content.

Images