JP2013088477A - Speech recognition system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a speech recognition system which shortens the time required for acquisition of a recognition result when performing distributed speech recognition processing in a client and a server and is capable of reducing the scale and processing burden of the client which relate to speech recognition processing.SOLUTION: Speech recognition processing on user's speech inputted to an in-vehicle device 1 is performed by the in-vehicle device 1 or a server 2 connected via a network 3. The in-vehicle device 1 includes: a microphone 22; a speech recognition processing unit 100 which takes a plurality of preliminarily prepared words or sentences as the object to perform speech recognition processing; a distribution determination unit 102 which determines an input speech as a speech to be subjected to speech recognition processing in the speech recognition processing unit 100 or a speech to be subjected to speech recognition processing in the server 2; and a speech data transmission unit 56 which, in the case of a speech to be subjected to speech recognition processing in the server, transmits speech data to the server 2.

Description

  The present invention relates to a speech recognition system that performs speech recognition processing on input speech.

  2. Description of the Related Art Conventionally, a voice recognition system that includes a server and a client connected via a network, processes what can be processed by the client by the client, and transmits only those that can only be processed by the server to the server for processing. It is known (for example, refer to Patent Document 1). The determination of which client or server speech recognition engine to use is basically based on whether or not the client has a small vocabulary dictionary group including one or more of an acoustic model dictionary, a language model dictionary, and a word dictionary. Done. That is, when there is a small vocabulary dictionary group, the speech recognition process is performed by the client, and when there is no small vocabulary dictionary group, the speech recognition process is performed by the server. Further, when the voice recognition process is not performed as a result of the client performing the voice recognition process, the server performs the voice recognition process.

Japanese Patent Laying-Open No. 2005-249829 (page 8-11, FIG. 1-3)

Patent Document 1 described above includes the following specific examples of whether voice recognition processing for the input voice is performed by the client or the server when voice input is performed.
(1) If a small vocabulary dictionary group including any one or more of an acoustic model dictionary, a language model dictionary, and a word dictionary according to the language of each country is prepared, the voice recognition according to each language can be performed by the client.
(2) A group of small vocabulary dictionaries containing at least one of a field-specific acoustic model dictionary, language model dictionary, or word dictionary, depending on whether it is used in a hospital or restaurant reservation system, or in Internet stock trading. If prepared, a small vocabulary dictionary group including one or more of an acoustic model dictionary, a language model dictionary, and a word dictionary can be selected for each field that the user wants to use, and the speech recognition hit rate at the client Can also be increased.

  As can be seen from these descriptions, in the speech recognition system of Patent Document 1, for a specific application (specific language or field), a small vocabulary dictionary group corresponding to this application is prepared in the client, and speech recognition processing is performed. In the case where the client cannot recognize or for other uses, the server performs voice recognition processing.

  However, in such a speech recognition system, if recognition is not possible as a result of performing speech recognition processing at the client, then speech recognition processing by the server is performed thereafter, and in such a case, a final recognition result is obtained. There was a problem that it took a long time to be released. This problem occurs when voice recognition processing is performed on both the client and server. However, if the client performs voice recognition processing reliably, it is difficult to obtain good recognition results. It is necessary to improve the size of the vocabulary dictionary group and the performance of the speech recognition engine using the lexical dictionary group, which is contrary to the purpose of performing speech recognition processing distributed between the client and the server.

  The present invention was created in view of the above points, and its purpose is to reduce the time required until a recognition result is obtained when performing voice recognition processing distributed between a client and a server. Another object of the present invention is to provide a voice recognition system that can reduce the size and processing load of a client related to voice recognition processing.

  In order to solve the above-described problems, the speech recognition system according to the present invention performs speech recognition processing on a user's speech input at a client by the client or a server connected to the client via a network. . The client includes voice input means for inputting voice to be recognized, client-side voice recognition processing means for performing voice recognition processing on a plurality of words or sentences prepared in advance, and voice input by the voice input means. For the voice recognition processing in the client-side voice recognition processing means and the sorting means for sorting the voice recognition processing in the server, and the voice that is distributed based on the voice recognition processing in the server by the sorting means Client-side communication means for transmitting the data to the server. The server includes server-side communication means for receiving voice data sent from the client, and server-side voice recognition processing means for performing voice recognition processing using the voice data received by the server-side communication means. Specifically, the above-described client-side voice recognition processing unit specifies a reading of a plurality of words or sentences prepared in advance by performing voice recognition processing on the voice input by the voice input unit. doing.

  By performing voice recognition processing on the client side only for words and sentences prepared in advance, it is possible to accurately distribute voice recognition processing performed on the client side and voice recognition processing performed on the server side. It is possible to avoid performing the voice recognition processing on both the server side, and it is possible to shorten the time required until a recognition result is obtained. In addition, since it is only necessary for the client side to obtain a recognition result for words and sentences prepared in advance, it is possible to reduce the size and processing burden of the client regarding the speech recognition processing.

  The client described above is an in-vehicle device, and the client further includes an input processing unit that performs an operation instruction or information input to the in-vehicle device in accordance with a recognition result by the client-side speech recognition processing unit or the server-side speech recognition processing unit. It is desirable to provide. As a result, various inputs in the in-vehicle device can be performed by voice input using voice recognition processing, and the time required for the voice recognition processing at that time can be shortened, and the scale and processing load of the in-vehicle device can be reduced. .

  The client described above further includes an operation unit that accepts a manual operation by a user, and the input processing unit uses a recognition result by the client side voice recognition processing unit, a recognition result by the server side voice recognition processing unit, and an operation unit. It is desirable to perform an operation instruction or information input to the in-vehicle device in accordance with any manual operation. As a result, when performing various inputs in the in-vehicle device, voice input using voice recognition processing and input by manual operation using operation means can be properly used as necessary, and operability can be improved. Become.

  In addition, it is desirable that the above-described distribution unit distributes to the voice recognition processing in the client side voice recognition processing unit instead of the voice recognition processing in the server when connection to the server cannot be performed via the network. Thereby, even if it is a case where connection with a server cannot be performed for some reason, it becomes possible to perform an operation instruction and information input using voice recognition processing.

  In addition, when the input processing unit described above cannot be connected to the server via the network, the operation instruction to the in-vehicle device is performed according to the manual operation using the operation unit instead of the recognition result by the server side voice recognition processing unit. Alternatively, it is desirable to input information. As a result, even if the connection with the server cannot be established due to some cause, it is possible to perform operation instructions and information input using the operation means.

  In addition, when a plurality of words or sentences that are targets of operation instructions or information input to the on-vehicle device described above are known, the sorting unit sorts the voice recognition processing by the client side voice recognition processing unit, and the client side voice It is desirable that the recognition processing means selects a word corresponding to the voice recognition result from a plurality of words or sentences by performing voice recognition processing on the voice input by the voice input means. Moreover, it is desirable that the client side speech recognition processing means described above has a speech recognition dictionary for a plurality of known words or sentences. As a result, the client-side speech recognition processing means can reliably extract the one corresponding to the input speech from words and sentences prepared in advance.

  Moreover, it is desirable that the plurality of words or sentences described above are a plurality of operation commands for instructing operation to the in-vehicle device. As a result, it is possible to perform voice recognition processing on the in-vehicle device side for operation instructions for the in-vehicle device, quickly determine the contents of the instruction, and reflect them in the operation of the in-vehicle device.

  In addition, the above-described in-vehicle device has a phone number included in the phone book data when a mobile phone having a built-in phone book data including a phone number of a call destination and a name corresponding to each phone number is connected. It is desirable to operate as a hands-free telephone system for making a call using a mobile telephone, and the plurality of words or sentences are preferably at least one of a telephone number and a name. The above-described client-side voice recognition processing means creates a voice recognition dictionary that creates a voice recognition dictionary corresponding to the reading of at least one of the phone number and name included in the phone book data when a mobile phone is connected. It is desirable to have a means. Thus, the present invention can be applied to input of a telephone number and a name in a hands-free telephone system as an in-vehicle device.

  The above-described in-vehicle device operates as an audio device that selectively reproduces a plurality of songs, and a plurality of words or sentences have a song name, album name, and artist name corresponding to each of the plurality of songs. At least one is desirable. The client-side voice recognition processing means described above preferably includes voice recognition dictionary creation means for creating a voice recognition dictionary corresponding to at least one reading of a song name, album name, and artist name. As a result, the present invention can be applied to the input of a song name, album name, and artist name in an audio device as an in-vehicle device.

  The on-vehicle device described above further includes document creation means for creating a document to be transmitted via the network, and when performing text input necessary for document creation based on the voice input by the voice input means, sorting is performed. It is desirable that the means is distributed to the voice recognition processing in the server side voice recognition processing means. The on-vehicle device described above further includes a recognition result acquisition unit that acquires a recognition result by the server side voice recognition processing unit transmitted from the server, and the document creation unit converts the recognition result acquired by the recognition result income unit into the document It is desirable to use it as the text necessary to create Thereby, it becomes possible to apply this invention to the text input at the time of creating documents, such as e-mail creation, in an in-vehicle device.

  The on-vehicle device described above further includes facility information display means for displaying the detailed information of the specific facility, and the input of the specific facility that is the target of the detailed information display by the facility information display means is input to the voice input by the voice input means. When performing based on this, it is desirable that the distribution unit distributes to the voice recognition processing in the server side voice recognition processing unit. The on-vehicle apparatus described above further includes a recognition result obtained by the server-side voice recognition means transmitted from the server, or a recognition result acquisition means for obtaining detailed information searched using the recognition result, and includes facility information display means. It is desirable to display the detailed information searched using the recognition result acquired by the recognition result acquisition unit or the detailed information acquired by the recognition result acquisition unit. Thereby, when displaying the detailed information of a specific facility in a vehicle-mounted apparatus, it becomes possible to apply this invention to the input of the specific facility used as a display target.

  Moreover, it is desirable that the voice input means described above is a microphone. As a result, the user can appropriately speak the voice recognition process on the client side or the server side only by speaking to the microphone provided in the client, and can acquire the recognition result in a short time.

  In addition, a speech switch that can be operated by the user when speaking to the microphone described above is further provided, and the distribution unit distributes the voice of the user collected by the microphone after the speech switch is operated. It is desirable to do. As a result, the input timing of the speech to be subjected to speech recognition processing becomes clear, and the processing procedure can be simplified and the recognition accuracy can be improved.

  In addition, a plurality of servers each including server-side voice recognition processing means can be connected to the client, and the sorting means selects one of the plurality of servers when sorting is performed on the server as performing voice recognition processing. It is desirable to do. Thereby, it is possible to request the voice recognition processing by properly using a plurality of servers having different fields of expertise, and the recognition accuracy when requesting the server can be improved.

It is a figure showing the whole voice recognition system composition of one embodiment. It is a figure which shows the detailed structure of a vehicle-mounted apparatus. It is a flowchart which shows the operation | movement procedure until a user speaks and the content is reflected on operation etc. of a vehicle-mounted apparatus. It is a flowchart which shows the operation | movement procedure of the modification in case a connection with a server cannot be performed. It is a flowchart which shows the operation | movement procedure of the other modified example when a connection with a server cannot be performed. It is a flowchart which shows the operation | movement procedure which reads a telephone directory data at the time of a mobile telephone connection, and registers a speech recognition dictionary. It is a flowchart which shows the operation | movement procedure which reads the attached information of a content list when a USB memory is connected, and registers a speech recognition dictionary. It is a figure which shows the whole structure of the speech recognition system of a modification. It is a flowchart which shows the operation | movement procedure of the modification in the case of using a some server properly.

  Hereinafter, a speech recognition system according to an embodiment to which the present invention is applied will be described with reference to the drawings. FIG. 1 is a diagram illustrating an overall configuration of a speech recognition system according to an embodiment. The voice recognition system according to the present embodiment includes an in-vehicle device 1 and a server 2. The in-vehicle device 1 has functions such as a navigation device and an audio device, and is mounted on the vehicle. The server 2 is provided outside the vehicle and is connected to the in-vehicle device 1 via a predetermined network 3. The network 3 is, for example, the Internet, and is connected via a mobile phone as a mobile phone connected to the in-vehicle device 1 and a base station (both not shown). Note that the connection of the in-vehicle device 1 to the network 3 is not necessarily performed via a mobile phone, and a wireless LAN communication device and an access point (both not shown) connected to (or built in) the in-vehicle device 1. You may make it connect via. When the wireless LAN can be connected to the network 3, the wireless LAN is connected. When the wireless LAN is not possible (such as when there is no access point nearby), the mobile phone is connected. You may do it.

  The in-vehicle device 1 includes a voice recognition processing unit 100 and a sorting determination unit 102, and the server 2 includes a voice recognition processing unit 200. In the voice recognition system according to the present embodiment, the voice recognition processing unit 100 in the in-vehicle device 1 or the in-vehicle device 1 is connected to the user's voice input in the in-vehicle device 1 as a client via the network 3. The voice recognition processing by the voice recognition processing unit 200 in the server 2 is performed, and the distribution determination unit 102 determines which of the in-vehicle device 1 and the server 2 performs the voice recognition processing.

  FIG. 2 is a diagram illustrating a detailed configuration of the in-vehicle device 1. As shown in FIG. 2, the in-vehicle device 1 includes a navigation processing unit 10, a TV tuner processing unit 14, a radio tuner processing unit 16, a voice input processing unit 20, a voice recognition processing unit 100, an operation unit 40, and a speech switch (SW). 42, input control unit 44, control unit 50, display processing unit 60, display device 62, digital-analog converter (D / A) 64, speaker 66, hard disk device (HDD) 70, USB (Universal Serial Bus) interface unit (USB I / F) 80 and 82 are provided.

  The navigation processing unit 10 performs a navigation operation for guiding the traveling of the vehicle on which the in-vehicle device 1 is mounted, using the map data stored in the hard disk device 70. Used in conjunction with the GPS device 12 for detecting the position of the host vehicle, the navigation operation for guiding the running of the vehicle searches for and displays peripheral facilities and POIs (Point Of Interest) in addition to map display and route search / guidance. Etc. are included. The vehicle position detection may be performed using a combination of autonomous navigation sensors such as a gyro sensor and a vehicle speed sensor in addition to the GPS 12.

  The TV tuner processing unit 14 receives a broadcast signal such as terrestrial digital broadcast and performs processing for reproducing video and audio. The radio tuner processing unit 16 receives radio broadcast signals and performs processing for reproducing sound.

  The voice input processing unit 20 performs input processing of the voice of the user (speaker) collected by the microphone 22. Specifically, the voice input processing unit 20 includes an analog-digital converter (A / D) 24 and a compression processing unit 26. The analog-digital converter 24 converts the output signal of the microphone 22 into digital audio data. The compression processing unit 26 compresses the audio data output from the analog-digital converter 24.

  The voice recognition processing unit 100 is for performing voice recognition processing on the voice collected by the microphone 22, and includes a voice recognition dictionary creation unit 32, a voice recognition dictionary 34, and a voice recognition unit 36. The voice recognition dictionary creation unit 32 is configured to provide a telephone number to be called when the in-vehicle device 1 operates as a hands-free telephone system and a name (including an individual name) corresponding to each telephone number. A voice recognition dictionary 34 corresponding to the voice that reads out the music title, album name, artist name, and the like to be reproduced when operating as an apparatus is created. The voice recognition dictionary 34 is created at a predetermined timing, and a specific example thereof will be described later. The voice recognition dictionary 34 is used for performing voice recognition processing on a plurality of known words or sentences, and the plurality of words or sentences are generated at a predetermined timing, such as the above-described telephone number or song name. In addition to what is performed, an operation command for instructing the in-vehicle device 1 to be operated is included. The voice recognition unit 36 performs voice recognition processing on the voice of the user collected by the microphone 22 using the voice recognition dictionary 34, and specifies the content (character string) of the voice uttered by the user.

  The operation unit 40 is for accepting a manual operation by the user with respect to the in-vehicle device 1 and includes various operation keys, operation switches, operation knobs, and the like. Further, when various operation screens and input screens are displayed on the display device 62, the user directly points a part of these operation screens and input screens with a finger or the like, so that display items of the operation screens and input screens are displayed. A touch panel that detects the position of the finger pointed to is a part of the operation unit 40 in order to enable an operation using such an operation screen or an input screen. As provided. Instead of using the touch panel, a part of the operation screen or the input screen may be selected according to a user instruction using a remote control unit or the like. The utterance switch 42 is used by the user to indicate the utterance timing when the user utters toward the microphone 22. The input control unit 44 monitors the operation unit 40 and the utterance switch 42 and determines the content of these operations.

  The control unit 50 controls the entire vehicle-mounted device 1 and operates as an audio device or a hands-free telephone system. The control unit 50 is realized by the CPU executing an operation program stored in a ROM or RAM. In FIG. 1, the navigation processing unit 10, the TV tuner processing unit 14, the radio tuner processing unit 16, the voice recognition processing unit 100, and the like are provided separately from the control unit 50, but some of these functions are provided. You may make it implement | achieve by the control part 50. FIG. Details of the control unit 50 will be described later.

  The display processing unit 60 outputs video signals for displaying various operation screens and input screens, video screens corresponding to broadcast signals received by the TV tuner processing unit 14, and displays these various screens on the display device 62. . The digital-analog converter 64 converts audio data when the in-vehicle device 1 operates as a hands-free telephone system into an analog audio signal and outputs it from the speaker 66, and also when the in-vehicle device 1 operates as an audio device. Audio data (music data) is converted into an analog audio signal and output from the speaker 66. In practice, an amplifier for amplifying a signal is connected between the digital-analog converter 64 and the speaker 66, but this amplifier is omitted in FIG. Further, the combination of the digital-analog converter 64 and the speaker 66 is provided for the number of reproduction channels, but only one set is shown in FIG.

  The hard disk device 70 stores map data and peripheral facility / POI search data used for the navigation operation by the navigation processing unit 10, content list used for reproduction operation in the audio device, phone book data used in the hands-free telephone system, and the like. Here, the content list includes the folder structure, file structure, and file attributes of content data (music data). This file attribute includes attached information corresponding to each song, specifically, the name of the artist who sings or performs the song, the album name if there is an album containing the song, and the name of this song (Song name). In addition, the phone book data includes a pre-registered phone number and a name corresponding to each phone number (in the case of an individual, the name and nickname of the individual, such as the company name and its name in the case of a company or other organization) Abbreviations, etc.).

  The USB interface units 80 and 82 are for inputting / outputting signals to / from a mobile phone 90 or a USB memory 92 as an external storage medium via a USB cable. included. Music data is recorded in the USB memory 92.

  Next, details of the control unit 50 will be described. As shown in FIG. 2, the control unit 50 includes a telephone directory acquisition unit 51, a telephone processing unit 52, a content list creation unit 53, an AV processing unit 54, an Internet processing unit 55, a distribution determination unit 102, an audio data transmission unit 56, A recognition result acquisition unit 57 and an input processing unit 58 are included.

  The phone book acquisition unit 51 reads and acquires the phone book data registered in the mobile phone 90 connected to either of the USB interface units 80 and 82. The acquired telephone directory data is stored in the hard disk device 70, for example. In this phone book data, the “phone number” to be called, the “name” such as the personal name or company name corresponding to each phone number, and the e-mail address if the e-mail address is known. Is included. In the present embodiment, the following description will be given on the assumption that the mobile phone 90 is connected to one USB interface unit 80 and the USB memory 92 storing music data is connected to the other USB interface unit 82.

  The telephone processing unit 52 selects the telephone number included in the telephone directory data acquired by the telephone directory acquisition unit 51 or when the user directly inputs the telephone number using the operation unit 40. A calling process for making a call using the mobile phone 90 is performed on the telephone number. In addition, after the telephone line is connected to the other party, the telephone processing unit 52 transmits the voice of the speaker collected by the microphone 22 to the other party and also transmits the voice of the other party to the speaker 66. Process to output from. In this manner, a hands-free telephone system using the mobile phone 90 is realized by the telephone processing unit 52.

  The content list creation unit 53 analyzes the content data (music data) recorded in the USB memory 92 as the recording medium from which the connection is detected, and creates a content list based on the analysis result. As described above, the content list includes the folder structure, file structure, and file attributes (artist name, album name, song name, etc.) of content data. The created content list is stored in the hard disk device 70, for example. The creation of the content list is performed, for example, at the timing when the USB memory 92 is connected to the USB interface unit 82.

  The AV processing unit 54 reads out music data in a predetermined format stored in the USB memory 92, performs demodulation processing, and converts it into music data (for example, PCM data) in a format input to the digital-analog converter 64. Play music. Further, the AV processing unit 54 creates a playback menu screen for the user to select a playback location of the music, change the volume, etc., during the playback operation. The reproduction menu screen is displayed on the display device 62 via the display processing unit 60.

  The Internet processing unit 55 performs processing necessary for using various services via the Internet. Specifically, the Internet processing unit 55 has functions of a web browser and mail software, and in response to a user's instruction and input, browsing of a web page, creation and transmission / reception of an email, SNS (social)・ Networking service) View and input screens.

  The distribution determination unit 102 performs voice recognition processing in the voice recognition processing unit 100 on the client side (in-vehicle device 1) and performs voice recognition processing in the voice recognition processing unit 200 of the server 2 on the voice collected by the microphone 22. Sort what to do. The voice data transmission unit 56 transmits the voice data collected by the microphone 22 and compressed by the compression processing unit 26 to the server 2 when requesting the voice recognition processing unit 200 of the server 2 to perform voice recognition processing. I do. The recognition result acquisition unit 57 receives the recognition result when the result (recognition result) of the voice recognition process is sent back from the server 2. The server 2 receives voice data sent from the in-vehicle device 1 and inputs it to the voice recognition processing unit 200, and obtains a recognition result by the voice recognition processing unit 200 and sends it back to the in-vehicle device 1. A communication control unit 202 is provided.

  The input processing unit 58 is an operation instruction to the in-vehicle device 1 among the recognition result by the voice recognition processing unit 100, the recognition result by the voice recognition processing unit 200 acquired from the server 2, and the operation content using the operation unit 40. Or it selects as the content of information input. Specific examples of these selections will be described later.

  The microphone 22 described above is a voice input unit, the voice recognition processing unit 100 is a client side voice recognition processing unit, the distribution determination unit 102 is a distribution unit, the voice data transmission unit 52 is a client side communication unit, and the communication control unit 200 is The speech recognition processing unit 200 corresponds to the server-side communication means and the server-side speech recognition processing means. The input processing unit 58 is an input processing unit, the operation unit 40 is an operation unit, the Internet processing unit 55 is a document creation unit, the recognition result acquisition unit 57 is a recognition result acquisition unit, and the navigation processing unit 10 is a facility information display. Each means corresponds.

  The voice recognition system according to the present embodiment has such a configuration. Next, an operation in which a user speaks toward the microphone 22 and performs voice recognition processing on the spoken voice will be described.

  FIG. 3 is a flowchart showing an operation procedure until the user speaks and the contents are reflected in the operation of the in-vehicle device 1. The voice recognition unit 36 determines whether or not the utterance switch 42 is turned on (step 100). If the utterance switch 42 is not turned on, a negative determination is made, and this determination is repeated.

  When the speech switch 42 is turned on, an affirmative determination is made in the determination of step 100. Next, the distribution determination unit 102 analyzes the input mode based on the display content at that time (step 102). For example, it corresponds to an input mode in which a menu screen or the like for selecting a corresponding item is displayed (this input mode is referred to as “corresponding item selection mode”), or a text box is included to enter a text input state. It is determined whether the input mode is applicable (this input mode is referred to as “text input mode”).

  By the way, in this embodiment, it is assumed that various operation instructions and information input are performed by voice input. In addition, the speech recognition process is performed on the input speech to identify the content, but a word or sentence for which a speech dictionary corresponding to one-to-one correspondence is prepared in advance or the speech recognition process is supported. The sorting determination unit 102 sorts whether or not an unspecified word or sentence for which a speech dictionary is not prepared in advance is to be subjected to speech recognition processing.

  More specifically, when the input mode is “corresponding item selection mode”, a word or sentence for which a speech dictionary corresponding to one-to-one correspondence is prepared in advance is subjected to speech recognition processing. The voice recognition process at this time is performed by the voice recognition processing unit 100 of the in-vehicle device 1. On the other hand, when the input mode is “text input mode”, an unspecified word or sentence for which a corresponding speech dictionary is not prepared in advance is targeted for speech recognition processing. Is performed by the voice recognition processing unit 200 of the server 2.

  The distribution determination unit 102 determines whether or not the analyzed input mode is a corresponding item selection mode (step 104). If the current input mode is the corresponding item selection mode, a positive determination is made. Next, the voice recognition processing unit 100 built in the in-vehicle device 1 performs voice recognition processing on the user's voice collected by the microphone 22 (step 106). Further, the input processing unit 58 uses the voice recognition result as an operation instruction or information input content corresponding to the display content at that time to execute an operation or input to the in-vehicle device 1 (step 108).

  For example, when a map image is being displayed during the navigation operation by the navigation processing unit 10, when a change in display scale is instructed when “shukusaku” is input as a voice, or when “mokutekichi” is input as a voice Instructing the user to set a destination is prepared in advance, and a voice recognition dictionary 34 for voice recognition of “shukusaku”, “mokutekichi”, and the like is prepared. The distribution determination unit 102 determines that the input mode is “corresponding item selection mode” when the display content at that time is “map image”, and the speech recognition unit 36 performs the processing for the input speech “Shukusaku” and the like. Speech recognition using the speech recognition dictionary 34 is performed, and a character string “shukusaku” or the like can be obtained as a recognition result. In response to this recognition result, the navigation processing unit 10 starts a process of changing the display scale of the map image being displayed.

  In addition, when a channel selection screen for selecting a broadcast station to be received is displayed during the reception operation by the TV tuner processing unit 14, the selection to this broadcast station is made when a voice is input as “XX TV”. An instruction to switch stations is determined in advance, and a voice recognition dictionary 34 for voice recognition such as “XX TV” is prepared. The distribution determination unit 102 determines that the input mode is “corresponding item selection mode” when the display content at that time is “channel selection screen”, and the voice recognition unit 36 inputs the input voice “XX TV” or the like. The voice recognition using the voice recognition dictionary 34 is performed for the character string, and the character string “XX TV” or the like can be obtained as a recognition result. Upon receiving this recognition result, the TV tuner processing unit 14 changes the channel selection to OO TV or the like.

  On the other hand, if the current input mode is the text input mode, a negative determination is made in the determination in step 104. Next, the voice data transmission unit 56 transmits the voice data input from the microphone 22 and compressed by the compression processing unit 26 to the server 2 via the network 3, and the voice recognition processing unit in the server 2. A voice recognition process 200 is requested (step 110). Thereafter, the recognition result acquisition unit 57 determines whether or not the recognition result sent back from the server 2 has been received (step 112). If not received, a negative determination is made and this determination is repeated. If a recognition result is received, an affirmative determination is made in the determination of step 112. Next, the input processing unit 58 uses the voice recognition result received from the server 2 as an operation instruction or information input content corresponding to the display content at that time to execute an operation or input to the in-vehicle device 1 (step 108). .

  For example, when the mail creation screen is displayed during the mail creation operation by the Internet processing unit 55 and the input position indicates the mail body, the sorting determination unit 102 determines that the “text input mode” is set, and The speech recognition processing unit 200 in the server 2 is requested to perform speech recognition (natural language recognition) on the input speech indicating the input text. When the recognition result (character string corresponding to the input text) is sent back, the Internet processing unit 55 creates a mail by inputting the character string as the recognition result in the mail body.

  As described above, in the speech recognition system according to the present embodiment, the speech recognition processing performed on the in-vehicle device 1 side and the server 2 are performed by performing the speech recognition processing on the in-vehicle device 1 side limited to words and sentences prepared in advance. Since the voice recognition processing performed on the side can be accurately distributed, it is possible to avoid performing the voice recognition processing on both the in-vehicle device 1 side and the server 2 side, and shorten the time required until the recognition result is obtained. can do. In addition, since it is sufficient that the recognition result is obtained only for words and sentences prepared in advance on the in-vehicle device 1 side, it is possible to reduce the scale and processing load of the in-vehicle device 1 related to the speech recognition processing.

  In addition, by providing the operation unit 40 in the in-vehicle device 1, when performing various inputs in the in-vehicle device 1, voice input using voice recognition processing and input by manual operation using the operation unit 40 are performed as necessary. The operability can be improved.

  In addition, when a plurality of words or sentences that are targets of operation instructions or information input to the in-vehicle device 1 are known, the distribution determination unit 102 performs the distribution for performing the voice recognition process in the in-vehicle device 1. A voice recognition dictionary 34 corresponding to a known word or sentence is provided in the in-vehicle device 1 to perform voice recognition processing, whereby the in-vehicle device 1 can reliably perform voice recognition processing for input speech.

  In addition, by using the plurality of known words or sentences described above as a plurality of operation commands for instructing the in-vehicle device 1, voice recognition processing on the in-vehicle device 1 side is performed for the operation instructions for the in-vehicle device 1. It is possible to quickly determine the content of the instruction and reflect it in the operation of the in-vehicle device 1.

  By the way, the operation procedure shown in FIG. 3 is based on the assumption that the server 2 can be connected to the server 2 at any time. However, when the vehicle on which the in-vehicle device 1 is mounted is traveling in a place where the radio waves of the mobile phone 90 do not reach or when the vehicle is parked in such a place, the server 2 may not be connected. For example, connection with the server 2 is often impossible when traveling in a mountainous area where the base station of the mobile phone 90 does not exist or while parked or traveling in a long tunnel.

  FIG. 4 is a flowchart showing the operation procedure of the modified example when connection to the server 2 cannot be made. The operation procedure shown in FIG. 4 is different from the operation procedure shown in FIG. 3 in that the operation of step 109 is added before the operation of step 110.

  In step 109, if the current input mode is the text input mode and a negative determination is made in the determination in step 104, the voice data transmission unit 56 determines whether or not the connection to the server 2 has been established. . If the connection has been established, an affirmative determination is made, and the operation proceeds to a voice recognition processing request operation by the voice recognition processing unit 200 in the server 2 in step 110.

  Further, if the connection of the server 2 is difficult (including the case where the cellular phone 90 is not connected or malfunctions in addition to the case where the radio wave condition is bad), a negative determination is made in the determination of step 109. In this case, the process proceeds to step 106, and the process proceeds to a voice recognition process using the voice recognition processing unit 100 built in the in-vehicle device 1. In the text input mode, since the input word or sentence is not known in advance, the speech recognition unit 36 performs speech recognition processing for each word of the user's speech. Processing for specifying the content is performed.

  As described above, since the voice recognition processing is performed in the in-vehicle device 1 when the connection with the server 2 is not possible, it is possible to prevent the processing from being interrupted due to the inability to connect to the server 2. .

  FIG. 5 is a flowchart showing the operation procedure of another modified example when connection to the server 2 is not possible. The operation procedure shown in FIG. 5 is different from the operation procedure shown in FIG. 4 in that an operation in step 111 is added as an operation after a negative determination is made in the determination in step 109.

  In this step 111, the connection with the server 2 cannot be established, and after a negative determination is made in the determination in step 109, an input operation using the operation unit 40 is performed. Instead of the voice recognition result received from the server 2, the input processing unit 58 uses the operation content of the operation unit 40 as an operation instruction or information input content corresponding to the display content at that time, and performs an operation or input to the in-vehicle device 1. Is executed (step 108).

  As described above, when the server 2 cannot be connected, a manual operation of the user using the operation unit 40 is performed instead of the voice recognition processing by the server 2, so that the server 2 cannot be connected. Can prevent the processing from being interrupted.

  Further, in the case where the speech recognition processing unit 100 in the in-vehicle device 1 performs speech recognition processing corresponding to the above-described corresponding item selection mode, the premise is that the words and sentences that are subject to speech recognition are known. It is necessary to provide a speech recognition dictionary 34 corresponding to these. For example, when the in-vehicle device 1 is used as an audio device, a navigation device, or the like, when the operation command (operation instruction) is performed by voice, a voice recognition dictionary 34 corresponding to a word or sentence as each operation command is created in advance. Just keep it.

  On the other hand, when the in-vehicle device 1 is used as a hands-free telephone system and the name and telephone number of a call destination are input by voice, or the in-car device 1 is used as an audio device and the artist name, album name, and song name are spoken. In the case of input or the like, since the content of voice input differs for each in-vehicle device 1 or for each connected mobile phone 90 or USB memory 92, the voice recognition dictionary 34 corresponding to a word or sentence that is a target of voice recognition. Need to be created as needed.

  FIG. 6 is a flowchart showing an operation procedure for reading the phone book data and registering the voice recognition dictionary 34 when the mobile phone 90 is connected. The phone book acquisition unit 51 determines whether or not the mobile phone 90 is connected (step 200), and makes a negative determination until it is connected and repeats this determination. If the mobile phone 90 is connected, an affirmative determination is made in the determination of step 200.

  Next, the phone book acquisition unit 51 reads the phone book data stored in the mobile phone 90 (step 202). This telephone directory data includes information such as an address in addition to the telephone number and name to be called. The telephone directory acquisition unit 51 extracts a telephone number and name from the read telephone directory data (step 204).

  Next, the speech recognition dictionary creation unit 32 in the speech recognition processing unit 100 creates a speech recognition dictionary corresponding to the extracted telephone number and name (step 206), and the speech recognition dictionary used by the speech recognition unit 36. 34 is registered (step 208). For example, the speech recognition dictionary creation unit 32 performs GTP (Grapheme To Phoneme, grapheme-phoneme conversion) processing on each extracted character string of the phone number and name, and obtains “reading information” of the character string. After the creation, a dynamic recognition dictionary for speech recognition processing is created from the read information. For example, a TTS (Text-to-Speech) process is performed on the read information to generate a speech waveform, and a feature recognition for speech recognition processing is performed on the speech waveform to create a dynamic recognition dictionary. . The created recognition dictionary is registered as the speech recognition dictionary 34.

  As described above, when the mobile phone 90 is connected, the phone book data is read out to extract the phone number and name, and a dynamic recognition dictionary is created. The call destination of the hands-free telephone system can be reliably determined by the recognition process.

  FIG. 7 is a flowchart showing an operation procedure for reading the attached information of the content list and registering the voice recognition dictionary 34 when the USB memory 92 is connected. The content list creation unit 53 determines whether or not the USB memory 92 is connected (step 300), makes a negative determination until it is connected, and repeats this determination. If the USB memory 92 is connected, an affirmative determination is made in the determination in step 300.

  Next, the content list creation unit 53 analyzes the content data recorded in the USB memory 92 (step 302). A content list is created based on the analysis result. This content list includes the folder structure, file structure, and file attributes (artist name, album name, music name, etc.) of the content data. The content list creation unit 53 extracts the artist name, album name, and song name from the content list created based on the analysis result (step 304).

  Next, the speech recognition dictionary creation unit 32 in the speech recognition processing unit 100 creates a speech recognition dictionary corresponding to the extracted artist name, album name, and song name (step 306), and is used by the speech recognition unit 36. Is registered as a voice recognition dictionary 34 (step 308). For example, the voice recognition dictionary creation unit 32 performs GTP processing on the extracted character strings of artist name, album name, and song name to create “reading information” of the character string, and then reads the reading information. To create a dynamic recognition dictionary for speech recognition processing. The created recognition dictionary is registered as the speech recognition dictionary 34.

  As described above, when the USB memory 92 is connected, the content data is read to extract the artist name, album name, and song name, and a dynamic recognition dictionary is created. The music to be reproduced in the audio apparatus can be reliably determined by the side voice recognition processing.

  Further, in the present embodiment described above, when the input mode is the “text input mode”, the server 2 is requested to perform the voice recognition processing. However, this request is distributed to a plurality of servers. You may do it. For example, when searching for peripheral facilities and POI (Point Of Interest) when using the in-vehicle device 1 as a navigation device, speech recognition processing (word recognition) is performed on words including facility names and the like. . On the other hand, when inputting by opening a text box when creating an e-mail, speech recognition processing (natural language recognition) is performed on a sentence rather than a word. Thus, a server suitable for each voice recognition process may be selected depending on the type of voice recognition process (whether word recognition or natural language recognition). Note that even if word recognition is generally used, there may be cases where the server that performs the speech recognition process suitable for the field of the word to be recognized is different. Therefore, the method of using a plurality of servers is divided according to the above-mentioned fields. In addition to the case, another division may be performed in consideration of the field and the like. Alternatively, a server with a low usage fee may be preferentially selected (as a precondition, it is necessary to hold information on the usage fee).

  FIG. 8 is a diagram illustrating an overall configuration of a modified speech recognition system. This voice recognition system includes an in-vehicle device 1 and a plurality of servers 2A and 2B. The server 2A includes a speech recognition processing unit 200A suitable for word recognition. Further, the server 2B includes a speech recognition processing unit 200B suitable for natural language recognition.

  FIG. 9 is a flowchart showing an operation procedure of a modified example in the case where a plurality of servers 2A and 2B are used properly. The operation procedure shown in FIG. 9 is different from the operation procedure shown in FIG. 3 in that the operation in step 110 is replaced with the operation in steps 110A and 110B. In the following, description will be given focusing on these replaced steps.

  If the current input mode is the text input mode and a negative determination is made in the determination in step 104, the voice data transmitting unit 56 next selects a server for requesting the voice recognition process (step 110A), and the microphone 22 is selected. Is sent to one of the selected servers 2A and 2B to request voice recognition processing (step 110B).

  For example, the server selection in step 110A is performed according to the type of voice recognition processing for the input voice. When performing word recognition, the server 2A is selected, and a speech recognition process is requested to the speech recognition processing unit 200A in the server 2A. Further, when performing natural language recognition, the server 2B is selected, and a speech recognition process is requested to the speech recognition processing unit 200B in the server 2B. As described above, a plurality of servers 2A and 2B having different fields of expertise and the like can be properly used to request voice recognition processing, and the recognition accuracy when requesting the server can be improved.

  In addition, this invention is not limited to the said embodiment, A various deformation | transformation implementation is possible within the range of the summary of this invention. For example, in the above-described embodiment, the vehicle-mounted device 1 has functions of a navigation device, an audio device, a hands-free telephone system, a television receiver, and a radio receiver. Voice recognition processing may be performed on the above.

  In the above-described embodiment, the “text input mode” is applied to the voice necessary for creating the e-mail. However, the “corresponding item selection mode” is applied to the input of the destination address. Good. For example, in step 204 shown in FIG. 6, an address is extracted from the read telephone directory data in addition to the telephone number and name, and in step 205, a speech recognition dictionary corresponding to the extracted telephone number, name, and address. Make sure to create. When the e-mail is created, if the input position on the e-mail creation screen is at the “destination” position where the e-mail destination address is entered and the address is entered by voice, this Voice recognition processing for input voice may be performed on the in-vehicle device 1 side. In addition to voice reading of the address itself, the voice input of the address may be voiced to read the name and telephone number corresponding to the address. Based on the phone book data acquired by the phone book acquisition unit 51, an address corresponding to the name (or telephone number) can be extracted one-to-one, and therefore, the extracted address can be assigned as an e-mail destination. it can.

  In the above-described embodiment, the audio data corresponding to the input voice is sent to the server 2 or the like, and the recognition result is sent back from the server 2 or the like toward the in-vehicle device 1. The result of performing the process may be sent back to the in-vehicle device 1. For example, when the in-vehicle device 1 is used as a navigation device and a voice input is performed for matters necessary for searching for peripheral facilities and POI, the voice recognition processing for the input voice is performed by the voice recognition processing unit 200 in the server 2 and the recognition is performed. A search processing unit (not shown) may be used to search for peripheral facilities using the result, and the search result may be sent back to the in-vehicle device 1.

  As described above, according to the present invention, the voice recognition processing on the in-vehicle device 1 side is performed on the server 2 side by performing the voice recognition processing on the in-vehicle device 1 side limited to words and sentences prepared in advance. Since the voice recognition processing to be performed can be accurately distributed, it is possible to avoid performing the voice recognition processing on both the in-vehicle device 1 side and the server 2 side, and to reduce the time required until the recognition result is obtained. Can do. In addition, since it is sufficient that the recognition result is obtained only for words and sentences prepared in advance on the in-vehicle device 1 side, it is possible to reduce the scale and processing load of the in-vehicle device 1 related to the speech recognition processing.

DESCRIPTION OF SYMBOLS 1 In-vehicle apparatus 2, 2A, 2B server 10 Navigation processing part 14 TV tuner processing part 16 Radio tuner processing part 20 Audio | voice input processing part 22 Microphone 24 Analog-digital converter (A / D)
26 compression processing unit 32 voice recognition dictionary creation unit 34 voice recognition dictionary 36 voice recognition unit 40 operation unit 42 speech switch (SW)
44 Input Control Unit 50 Control Unit 51 Telephone Book Acquisition Unit 52 Telephone Processing Unit 53 Content List Creation Unit 54 AV Processing Unit 55 Internet Processing Unit 56 Audio Data Transmission Unit 57 Recognition Result Acquisition Unit 58 Input Processing Unit 60 Display Processing Unit 62 Display Device 64 Digital-analog converter (D / A)
66 Speaker 70 Hard Disk Drive (HDD)
80, 82 USB interface (USB I / F)
DESCRIPTION OF SYMBOLS 90 Cellular phone 92 USB memory 100 Voice recognition processing part 102 Distribution determination part 200, 200A, 200B Voice recognition processing part 202 Communication control part

Claims (20)

A speech recognition system that performs speech recognition processing on a user's voice input at a client by a server connected to the client or the client via a network,
The client
Speech input means for inputting speech to be recognized;
Client-side speech recognition processing means for performing speech recognition processing on a plurality of words or sentences prepared in advance;
A sorting unit that sorts a voice input by the voice input unit between a voice recognition process performed by the client side voice recognition processing unit and a voice recognition process performed by the server;
A client-side communication unit that transmits voice data distributed as a result of performing voice recognition processing in the server by the distribution unit, to the server;
The server
Server-side communication means for receiving audio data sent from the client;
And a server-side voice recognition processing unit that performs voice recognition processing using voice data received by the server-side communication unit. In claim 1,
The client-side voice recognition processing means identifies a reading of any of the plurality of words or sentences prepared in advance by performing voice recognition processing on the voice input by the voice input means. Voice recognition system. In claim 2,
The client is an in-vehicle device,
The client further includes an input processing means for performing an operation instruction or information input to the in-vehicle device in accordance with a recognition result by the client side voice recognition processing means or the server side voice recognition processing means. system. In claim 3,
The client further comprises an operation means for accepting a manual operation by a user,
The input processing unit is configured to perform an operation instruction on the in-vehicle device or a recognition result by the client side voice recognition processing unit, a recognition result by the server side voice recognition processing unit, or a manual operation using the operation unit. A speech recognition system characterized by inputting information. In claim 3 or 4,
The voice recognition system according to claim 1, wherein when the connection unit cannot connect to the server via a network, the voice recognition process is performed by the client side voice recognition processing unit instead of the voice recognition process by the server. In claim 4,
When the input processing unit is unable to connect to the server via a network, the input processing unit performs an operation on the in-vehicle device according to a manual operation using the operation unit instead of the recognition result by the server side voice recognition processing unit. A speech recognition system characterized by giving instructions or inputting information. In any one of Claims 3-6,
When the plurality of words or sentences that are the target of operation instructions or information input to the in-vehicle device are known, the distribution unit distributes the voice recognition processing by the client side voice recognition processing unit, and the client side The voice recognition processing means performs voice recognition processing on the voice input by the voice input means, thereby selecting one corresponding to the voice recognition result from the plurality of words or sentences. Recognition system. In claim 7,
The client-side speech recognition processing means has a speech recognition dictionary for the plurality of known words or sentences. In claim 8,
The speech recognition system, wherein the plurality of words or sentences are a plurality of operation commands for instructing the in-vehicle device. In claim 8,
The in-vehicle device is connected to the phone number included in the phone book data when a mobile phone with a built-in phone book data including a phone number of a call destination and a name corresponding to each phone number is connected. On the other hand, it operates as a hands-free telephone system for making a call using the mobile phone,
The speech recognition system, wherein the plurality of words or sentences are at least one of the telephone number and the name. In claim 10,
The client side voice recognition processing means creates the voice recognition dictionary corresponding to the reading of at least one of the telephone number and the name included in the telephone directory data when the mobile phone is connected. A speech recognition system comprising a dictionary creating means. In claim 8,
The in-vehicle device operates as an audio device that selectively reproduces a plurality of music pieces,
The speech recognition system, wherein the plurality of words or sentences are at least one of a song name, an album name, and an artist name corresponding to each of the plurality of songs. In claim 12,
The client-side voice recognition processing means includes voice recognition dictionary creation means for creating the voice recognition dictionary corresponding to at least one reading of the song name, album name, and artist name. In any one of Claims 3-13,
The in-vehicle device further includes document creation means for creating a document to be transmitted via a network,
The voice is characterized in that when the text input necessary for creating the document is performed based on the voice inputted by the voice input means, the sorting means sorts the voice recognition processing in the server side voice recognition processing means. Recognition system. In claim 14,
The in-vehicle device further includes recognition result acquisition means for acquiring a recognition result by the server side voice recognition processing means transmitted from the server,
The speech recognition system, wherein the document creation means uses the recognition result acquired by the recognition result income means as text necessary for creation of the document. In any one of Claims 3-13,
The in-vehicle device further includes facility information display means for displaying detailed information of a specific facility,
When performing the input of the specific facility that is the target of detailed information display by the facility information display unit based on the voice input by the voice input unit, the distribution unit performs voice recognition in the server side voice recognition processing unit. A voice recognition system characterized by distribution to processing. In claim 16,
The in-vehicle device further includes a recognition result obtained by the server-side voice recognition means transmitted from the server, or a recognition result obtaining means for obtaining the detailed information searched using the recognition result,
The facility information display means displays the detailed information searched using the recognition result acquired by the recognition result acquisition means or the detailed information acquired by the recognition result acquisition means. system. In any one of Claims 1-17,
The voice recognition system, wherein the voice input means is a microphone. In claim 18,
A speech switch operable by a user when speaking toward the microphone;
The voice recognition system characterized in that the sorting means sorts the voice of the user collected by the microphone after the utterance switch is operated. In any one of Claims 1-19,
A plurality of the servers each including the server side voice recognition processing means can be connected to the client,
The distribution means selects one of the plurality of servers when performing distribution as performing speech recognition processing in the server.

Images