JP5037041B2 - On-vehicle voice recognition device and voice command registration method - Google Patents

Description

  The present invention relates to an in-vehicle voice recognition device and a voice command registration method, and in particular, based on voices such as words and phrases (hereinafter also referred to as “Voice Tags”) issued as commands by a user in a vehicle interior. The present invention relates to a technology adapted to perform accurate speech recognition.

  Recent vehicles are equipped with devices and devices for providing various services to users in the passenger compartment. Typical in-vehicle devices include a navigation device equipped with a function (route guidance function) for guiding the user to travel to a set destination without making a mistake, and various sources (radio receiver, CD player, TV) There are audio / video (A / V) devices that provide various entertainment such as audio (audio) information and video (video) information output from a receiver, a DVD player, and the like. These in-vehicle devices (apparatuses) are operated according to the operation instructions when the user gives a necessary instruction by operating the remote controller or the operation panel. The changed operating state of the device can be heard via a speaker (such as wireless headphones for a user at the rear seat) installed in the passenger compartment, or can be viewed through a screen of a display device such as an in-vehicle monitor.

  As described above, a required operation instruction can be input to each in-vehicle device based on a manual operation such as a remote control operation. However, recently, the control of the device can be performed only by voice input (speech) of the operation instruction. Devices equipped with a function that can be performed (voice recognition function) have also appeared. Such a voice recognition function is advantageous in terms of convenience for the user's operation, and is particularly useful for the driver in terms of safe driving.

  To realize the voice recognition function, a voice command recognition dictionary is required. This recognition dictionary should be controlled based on the voice of words or phrases that are the target of voice recognition, that is, voice recognition. A voice tag related to an operation instruction of the in-vehicle device or a voice command including the voice tag is registered. The voice tag is used, for example, when calling an address book in a navigation device. An example is shown in FIG.

  When registering a voice tag (see FIG. 9A), first, with the navigation function enabled, the “Address Book” screen 61 is displayed by remote control operation or touch operation on the screen, and then the screen. When an “Add Voice Tag” portion 62 on 61 is touched, guidance information (sub-screen) 63 for recording a voice tag is displayed. When the user touches the “Start” portion 64 on the screen 63 and utters a desired voice tag (sound to be called) within 2 seconds, for example, guidance information indicating that the voice tag is being recorded ( Child screen) 65 is displayed. When the “Finished” portion 66 on the screen 65 lights up, registration of the voice tag in the recognition dictionary is completed. For example, assume that the registered voice tag is [McDonald].

  When the voice tag (= [McDonald]) registered in this way is used for voice recognition processing (see FIG. 9B), first, a map screen around the vehicle position by a touch operation or the like on the screen. In this state, when the user utters “Go to [McDonald]”, voice recognition processing is executed based on this voice command (PCM data), and information such as a guidance route to the destination is displayed on the screen. 67. In the figure, CM is the vehicle position mark, GR is the guidance route, DS1 is information indicating the distance and time to the destination and its direction, and DS2 is information indicating the distance to the next guidance point on the guidance route. Show.

  In addition, when registering a voice tag in FIG. 9A, if the voice tag spoken by the user is data that cannot be used for speech recognition, the user is prompted to re-register as shown in FIG. 9C. Guidance information (warning screen) 68 is displayed. If necessary, the user touches the “Start” portion on the screen 68 to utter the voice tag again, and if not necessary, touches the “Cancel” portion to cancel the re-registration. Note that “data that cannot be used for speech recognition” here is basically a speech recognition engine because the user uttered in a very noisy environment (although it was registered for the time being). The voice data to be used is data when the effective level has not been reached. However, the warning screen 68 illustrated in FIG. 9C is not limited to such a case, and when the utterance itself cannot be detected (refer to the example of FIG. "Start" is also displayed when 2 seconds have passed without touching the portion 64).

As a technique related to the above prior art, for example, as described in Patent Document 1, in a speech recognition apparatus, a matching unit that performs matching between a speech pattern of input speech data and a standard speech pattern generated in advance The preliminary selection unit matches the input voice data with the full-band dictionary data analyzed by the full-band filter from the voice data in the dictionary generation unit and registered in the dictionary. The matching processing unit narrows down the candidate words, and the matching processing unit matches the narrowed candidate words with the band-by-band dictionary data analyzed by the band-by-band filter in the dictionary generation unit and registered in the dictionary. Candidate words whose similarity is greater than a predetermined threshold are output as recognition words. There are things you.
JP-A-6-301399

  As described above, in the conventional technique, a function has been realized in which a voice tag issued by a user is registered in a recognition dictionary and can be used for control of a navigation device or the like. However, in the conventional method, the function is registered in the recognition dictionary. Since voice recognition is performed on all the voice tags, there are inconveniences as described below.

  That is, the speech recognition engine calculates the degree of matching (also called “score”) between the voice tag (command) issued by the user and all the commands registered in the recognition dictionary, and the best match is obtained from the calculation result. A command having a high degree is determined as a command uttered by the user (voice recognition). At this time, there is no problem as long as the command with the highest degree of matching can be specified, but as the number of registered commands increases, the number of commands that are similar to “read” on the utterance increases. However, it is not always possible to specify one, and as a result, a command that does not match may be erroneously recognized. That is, in the conventional technology, when registering a voice tag, if the same or similar voice as a previously registered voice tag is registered, a plurality of identical or similar voice data is registered in the voice tag calling recognition dictionary. Therefore, there is a problem that the rate of erroneous recognition increases when voice recognition processing is performed when used for control of a navigation device or the like (that is, the recognition rate for voice commands decreases). .

  The present invention was created in view of the problems in the prior art, and improves the recognition rate for a registered voice command when the voice command is registered in a recognition dictionary and used to control an in-vehicle device such as a navigation device. It is an object of the present invention to provide a vehicle-mounted speech recognition device and a speech command registration method.

In order to solve the above-described problems of the prior art, according to one aspect of the present invention, a comparison is made between voice input means for inputting a command spoken by a user in a vehicle cabin and commands input via the voice input means. A plurality of recognition dictionaries in which commands that can be selected according to the operation state of the control target device are registered, and for determining whether the input commands should be registered in the recognition dictionary. Storage means storing a dictionary adapted to register the same command as that registered in the plurality of recognition dictionaries, and the recognition dictionary effective according to the operating state of the control target device a dictionary switching selection means for switching a comparison of commands said registered in any of the dictionaries stored in the storage means and the input command through the voice input means Voice recognition means for performing voice recognition based on the command, and the voice recognition means uses the discrimination dictionary only when the command is inputted via the voice input means. When the recognition score calculated based on the voice recognition is lower than a predetermined threshold, the recognition dictionary selected according to the operation state of the control target device and the determination An in-vehicle voice recognition device characterized by being registered in a dictionary is provided.

  According to the in-vehicle speech recognition device of the present invention, when registering a command (including a voice tag) issued by a user, speech recognition is performed on the command using only the discrimination dictionary, and a recognition score is based on the result. Is lower than the threshold value, the command is determined to be data that is not similar to any of the commands registered so far, and the command is registered in the recognition dictionary and the discrimination dictionary.

  In other words, it is determined whether the command to be registered (including the voice tag) is the same as or similar to that already registered in the discrimination dictionary, or is not similar (dissimilar). Only the command is registered. In other words, a command that is the same as or similar to a command that has already been registered is not registered.

  This eliminates the inconvenience seen in the prior art (confusingness in recognition caused by registration of voice data that is the same as or similar to previously registered commands (including voice tags)). It is possible to increase the recognition rate for the registered voice command.

According to another aspect of the present invention, the voice input means for inputting a command spoken by the user in the vehicle interior and the command input via the voice input means are used for comparison and collation, A plurality of recognition dictionaries in which commands that can be selected according to the operating state of the device are registered, and a determination dictionary for determining whether or not the input commands should be registered in the recognition dictionary. A storage unit storing a command adapted to register the same command as the registered command, a dictionary switching selection unit that switches the recognition dictionary that is valid according to an operation state of the control target device, and the voice A voice recognition that performs voice recognition based on comparison and collation between a command input via the input means and a command registered in any of the dictionaries stored in the storage means. In-vehicle speech recognition and means, when the command through the voice input means is input, executes the voice recognition for the determination dictionary only enable and the input command, the voice When the recognition score calculated based on recognition is lower than a predetermined threshold value, the command is registered in the recognition dictionary and the discrimination dictionary selected according to the operation state of the control target device, A voice command registration method is provided.

  Other structural features of the vehicle-mounted speech recognition apparatus according to the present invention and specific processing modes based thereon will be described in detail with reference to embodiments of the invention described later.

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

  FIG. 1 shows a configuration of an in-vehicle audio / video (A / V) navigation system incorporating an in-vehicle voice recognition device according to an embodiment of the present invention.

  As shown in the figure, the in-vehicle A / V / navigation system 40 is subject to control corresponding to the utterance content (command including a voice tag) based on the in-vehicle voice recognition device 10 according to the present embodiment and the voice recognition result. Equipment (in the illustrated example, radio receiver 1, DVD / CD player 2, TV receiver 4 and navigation unit 5) and a front seat for a user at the front seat to perform various setting operations on each control target equipment Operation unit (head unit (H / U)) 20, a rear seat operation unit 30 for a rear seat user to perform various setting operations on each control target device (excluding the navigation unit 5), front A seat display unit 25, an amplifier unit 26, a speaker 27, a rear seat display unit 31, and wireless headphones 32. It is provided. The in-vehicle voice recognition device 10, the control target devices 1 to 5, the front seat operation unit 20, the display units 25 and 31, and the amplifier unit 26 are mutually connected via a bus 6 such as an optical fiber provided as a transmission path. It is connected to the. In the example shown in the figure, only one speaker 27 is shown, but in reality, the required number in a predetermined position in the passenger compartment, for example, at least the vicinity of the left and right of the rear seats when the rear seats are one row. In total, four speakers 27 are installed in the vicinity of the left and right sides of the front seat. Similarly, only one unit (one) is shown for the operation unit 30 for the rear seat, the display unit 31 and the wireless headphones 32, but in reality, the required number (in accordance with the number of passengers in the rear seat) Number).

  The in-vehicle voice recognition device 10 according to the present embodiment includes a hard disk drive (HDD) 7 as a recording medium, a microphone 8, and a voice recognition unit 9. The microphone 8 is appropriately installed in the vicinity of the sun visor or the room mirror in front of the driver's seat in the vehicle interior, detects a command (including a voice tag) uttered by the user, and converts it into an analog voice signal corresponding to the sound pressure level. Is. A disk (not shown) driven by the HDD 7 stores map data used when the navigation function is executed and data used when the voice recognition function is executed in respective allocated storage areas. . The map data is divided into longitude and latitude widths of appropriate sizes according to each scale level (1/12500, 1/25000, 1 / 50,000, etc.), and can be used for various processes such as route search and map matching. Necessary road unit data, intersection unit data showing details of the intersection, data on various facilities (convenience store, gas station, super discount shop, etc.) (various information such as location, address, telephone number, genre, etc.) Contains. The contents of voice recognition data stored in the HDD 7 (part of its storage area) will be described later together with the internal configuration of the voice recognition unit 9.

  The front seat operation unit (H / U) 20 is installed in the form of an “operation panel” on the center console between the two seats so that the driver and the passenger in the front passenger seat can share the display. The unit 25 is disposed above the operation panel (H / U), and includes, for example, an LCD monitor. On the screen of the display unit 25, various kinds of video information output from the navigation unit 5 (a map around the vehicle position based on voice recognition, a guide route to the destination, etc.), DVD / CD player 2 and TV reception The video information output from the video source such as the machine 4 is displayed. On the other hand, the rear seat operation unit 30 is provided in the form of a “remote control” so that the user at the rear seat can easily operate, and is connected to the corresponding rear seat display unit 31 by infrared communication. The rear seat display unit 31 is installed, for example, in the rear part of the headrest of the front seat, and has an LCD monitor or the like that displays video information on the display screen, like the front display unit 25. The display unit 31 is connected to the corresponding wireless headphones 32 by infrared communication and RF communication. Instead of the wireless headphones 32, headphones with a jack may be used. In this case, the headphones are wired to the corresponding display unit 31 via a jack.

  Each of the control target devices 1 to 5 has, as a basic operation, data related to an operation instruction sent to the bus 6 from the front seat operation unit 20, the rear seat operation unit 30, or the voice recognition unit 10 (the “device” described later). Control signal ") is received, and its own operation state is set or changed based on the data related to the operation instruction, and data indicating the result (current operation state) is sent to the bus 6 as an audio / video signal. . For example, in the case of the radio receiver 1, in response to an operation instruction given from each operation unit 20, 30 or the voice recognition unit 10, an audio signal is generated by receiving and demodulating an FM broadcast or AM broadcast signal. This is converted into digital audio data and sent to the bus 6. In the case of the DVD / CD player 2, in response to an operation instruction given in the same manner, a signal recorded on the recording surface of the DVD selected by the user is read and the reproduced video data is sent to the bus 6. In the case of the navigation unit 5, in response to an operation instruction given in the same manner, a guidance route to the destination set by the user is searched, and data of the searched route is sent to the bus 6.

  The front seat operation unit 20 includes a control unit 21, an operation unit 22, a display unit 23, and a memory unit 24. Among these, the operation unit 22 selects operation keys for performing various setting operations on the control target devices 1 to 5, for example, a power key for power on / off and volume adjustment, and each device. Selection keys, shift keys and preset keys for performing predetermined operations and functions. The display unit 23 is arranged in the form of an LCD or the like on the operation panel (H / U), and based on data output from the control unit 21, various information, for example, FM / AM for the radio receiver 1 is used. For the DVD / CD player 2, the disc number and playback position (number of tracks, elapsed time, etc.) during CD performance are displayed. The memory unit 24 includes a nonvolatile memory such as a flash memory, and stores necessary information (data) based on the control from the control unit 21. For example, data indicating the operation state of the device at the time when the output operation of the audio / video signal from the selected device is stopped based on the operation instruction given from each operation unit 20, 30 or the voice recognition unit 10 is stored. The This data is stored for reference when necessary at the start of the next output operation. For example, this data may be a “source type” indicating which device (source) is used or an audio source. For example, “volume / sound quality” for instructing an adjustment value of volume and sound quality when listening to the sound, “detailed information for each device” for instructing a detailed operation state for each device, and the like are included.

  The control unit 21 is configured by a microcomputer or the like, and controls the entire system 40. Basically, on the basis of operation instructions given from the operation units 20 and 30 or the voice recognition unit 10, the audio / video data sent from the selected device via the bus 6 is acquired to obtain the audio / video information. Control is performed such as an operation for performing reproduction, an operation for displaying information indicating an operation state, an operation state, and the like on the display unit 23. In this case, the acquired audio data is sent to the amplifier unit 26 via the bus 6 by the control unit 21 and appropriately D / A converted, and the volume and sound quality are controlled and amplified, and then the speaker 27 is output as audio. The acquired video data is sent to the display unit 25 by the control unit 21 via the bus 6 and displayed as video information on the display screen.

  On the other hand, the rear seat operation unit (remote control) 30 is not particularly shown, but an operation unit having the same function as the operation unit 22 on the front side and a signal corresponding to an operation instruction input from the operation unit are infrared rays. An infrared transmission unit for transmitting to the display unit 31 by communication. The rear seat display unit 31 is not particularly shown, but is equivalent to the infrared communication unit for communicating control signals and data between the remote controller 30 and the wireless headphones 32, and the front side control unit 21. A control unit that performs control, a display unit that includes an LCD monitor or the like similar to the front-side display unit 25, and a memory unit similar to the front-side memory unit 24 are provided.

  Next, the configuration of the in-vehicle speech recognition device 10 according to the present embodiment will be described with reference to FIG.

  The on-vehicle speech recognition apparatus 10 according to the present embodiment includes an HDD 7 (part of its storage area), a microphone 8 and a speech recognition unit 9 as shown in the figure. The voice recognition unit 9 includes a voice input unit 11, a voice recognition processing unit 12, a dictionary switching selection unit 13, a voice reproduction processing unit 14, and a device control signal generation unit 15 as functional blocks.

  The HDD 7 stores, as voice recognition data, a command recognition dictionary for recognizing commands (including voice tags) spoken by the user, a discrimination dictionary JD that characterizes the present invention, and an acoustic model AM. ing. The acoustic model AM is well known to those skilled in the art. For example, a phoneme HMM set composed of phoneme HMMs (Hidden Markov Models) corresponding to phonemes is generated, and the phoneme HMMs of the phoneme HMM sets are combined to correspond to each syllable Can be created by generating an initial phoneme chain syllable HMM set of initial phoneme chain syllable HMMs and learning the initial phoneme chain syllable HMM set. The acoustic model AM is appropriately referred to when the voice recognition processing unit 12 performs voice recognition.

  For example, the command recognition dictionary is classified for each recognition dictionary including commands that can be selected in association with the operation state or operation instruction of each control target device 1 to 5 (in the illustrated example, three types of D1 to D3). Dictionary) and the dictionary identification number (ID = 1 to 3). In the illustrated example, the first command recognition dictionary D1 is assigned as a voice tag recognition dictionary.

  The determination dictionary JD is used when determining whether or not a voice tag issued by a user should be registered in the voice tag recognition dictionary D1 as described later. In this discrimination dictionary JD, at least the same voice tag as that registered in the voice tag recognition dictionary D1 is registered. Further, in the discrimination dictionary JD, commands supported by the navigation unit 5 as products (“Menu”, “Cancel”, “Map” in the example of FIG. 2) are registered as reserved words. By registering such reserved words in advance, when a voice tag is recognized as described later, it is determined whether the command is newly registered by the user or originally used as a navigation function. be able to.

  In the voice recognition unit 9, the voice input unit 11 appropriately amplifies a voice command (analog voice signal) issued by the user via the microphone 8, digitizes it, and outputs it to the voice recognition processing unit 12. The speech recognition processing unit 12 basically has a recognition dictionary (command recognition dictionaries D1 to D3 in the illustrated example) consisting of commands that can be selected in the operating state of each control target device 1 to 5 while referring to the acoustic model AM. ), The input voice command and each command included in the recognition dictionary are compared and collated, and the degree of matching (recognition score) is calculated. It is determined as a voice command issued. Further, as a process related to the present invention, the voice recognition processing unit 12 performs a recognition process using the discrimination dictionary JD when registering a voice tag, as will be described later, and determines the voice tag according to the recognition score. It is determined whether or not to be registered in the dictionary D1 for voice tag recognition.

  The dictionary switching selection unit 13 is operatively connected to each of the control target devices 1 to 5, and detects a change in these operation states, and from selectable commands corresponding to the operation states A recognition dictionary is selected. The voice reproduction processing unit 14 performs talkback reproduction of a recognition result of a command (including a voice tag) inputted by the user, and voice data for notifying a result of a voice operation on each of the control target devices 1 to 5. Is synthesized. The synthesized audio data is sent to the bus 6 (FIG. 1) and notified to the user from the speaker 27 via the amplifier unit 26. The device control signal generation unit 15 acquires the command determined by the voice recognition processing unit 12 and outputs a device control signal corresponding to the content of the command. The output device control signal is sent to the bus 6 (FIG. 1), and the corresponding control target device changes the operation state based on the device control signal.

  Hereinafter, with reference to FIG. 3 showing an example of processing related to voice tag registration based on switching selection between a command recognition dictionary and a discrimination dictionary performed in the in-vehicle speech recognition apparatus 10 (FIG. 2) according to the present embodiment. explain. In addition, a supplementary explanation will be given with reference to FIGS.

  First, as an initial state, it is assumed that no voice tag is registered in the voice tag recognition dictionary (command recognition dictionary D1) and the discrimination dictionary JD, and the discrimination dictionary JD has a command ( “Menu”, “Cancel”, “Map”) are registered as reserved words.

  In the first step S1 in this state, the voice recognition unit 9 determines whether voice tag voice data is detected from the microphone 8 via the voice input unit 11 (YES) or not (NO). If the determination result is YES, the process proceeds to step S2, and if the determination result is NO, the determination process is repeated until a voice tag is detected.

  In the next step S2, in the voice recognition unit 9, the dictionary switching selection unit 13 cooperates with the voice recognition processing unit 12 to select a command recognition dictionary (D1 to D3 in the example of FIG. 2) and a discrimination dictionary. Enable JD.

  In the next step S3, the voice recognition processing unit 12 refers to the dictionaries stored in the HDD 7 (in this case, the voice tag recognition dictionary D1 and the discrimination dictionary JD), and there are registered voice tags ( YES) or not (NO). If the determination result is YES, the process proceeds to step S4, and if the determination result is NO, the process proceeds to step S5.

  In step S5 (when the voice tag is not yet registered), the voice recognition processing unit 12 records the PCM data (voice data) of the detected voice tag and stores the voice tag in the recognition dictionary (dictionary D1 for voice tag recognition). ) And the registration dictionary JD. FIG. 4 shows an example. In the example shown in FIG. 4, a “Best Buy” command is registered as a voice tag in the voice tag recognition dictionary D1 and the discrimination dictionary JD ((b) and (c) in the figure. )reference). The screen 51 shown in FIG. 9A is the same as the “Address Book” screen 61 illustrated in FIG. When the registration of the voice tag is thus completed, the process returns to step S1 and the above processing is repeated.

  On the other hand, in step S4, the dictionary switching selection unit 13 cooperates with the voice recognition processing unit 12 to hold the ID of the currently effective recognition dictionary (ID = 1 to 3 in the example of FIG. 2). Then, all the dictionaries are invalidated (see FIG. 4C, FIG. 5B, and FIG. 7B).

  In the next step S6, the dictionary switching selection unit 13 cooperates with the voice recognition processing unit 12 to validate only the discrimination dictionary JD among the invalidated dictionaries (FIG. 5 (c), FIG. 7 (c)). The voice recognition processing unit 12 executes voice recognition for the detected voice tag using only the discrimination dictionary JD based on the recorded PCM data. FIGS. 5A and 7A show an example of the “detected voice tag” in this case. In the example of FIG. 5A, “My Home” is detected as the voice tag. In the example of (a), “Best Buy” is detected as a voice tag. Note that the screen 52 shown in FIG. 9A is equivalent to the screen 65 illustrated in FIG.

  In the next step S7, the voice recognition processing unit 12 determines whether the recognition score calculated based on the voice recognition is lower than a predetermined threshold (YES) or not (NO). If the determination result is yes, the process proceeds to step S8, and if the determination result is no, the process proceeds to step S9. Here, when the recognition score is lower than the threshold value, it can be determined that the detected voice tag is data that is not similar to any voice tag registered so far (unsimilar data). On the other hand, if the recognition score is higher than the threshold value, the detected voice tag may be determined to be the same or similar data (identical / similar data) to any of the registered voice tags. it can.

  In step S8 (when the detected voice tag is dissimilar data), the voice recognition processing unit 12 records the PCM data (voice data) of the detected voice tag, and the voice tag is used for the discrimination dictionary JD and the voice tag recognition. Registered in the dictionary D1. FIG. 6 shows an example. In the example shown in FIG. 6, a command “My Home” is newly registered as a voice tag (in FIG. 6, (b) voice tag recognition dictionary D1 ′, (c) (See JD ′ for determination). When the registration of the voice tag is thus completed, the process proceeds to step S10.

  On the other hand, in step S9 (when the detected voice tag is the same / similar data), the display unit (in this case, the front seat display unit 25 in this case) is controlled via the device control signal generator 15 based on the control from the voice recognition processor 12. ), Guidance information (warning screen) indicating that the voice tag is not registered in each dictionary (voice tag recognition dictionary D1 and discrimination dictionary JD) is displayed. FIG. 8A shows an example. In the example shown in the figure, guidance information (warning screen) 53 indicating that registration is not performed and re-registration is displayed because it is very similar to a voice tag registered in the past. Has been. When this warning screen is displayed, the process proceeds to step S10.

  In the last step S10, the dictionary switching selection unit 13 cooperates with the voice recognition processing unit 12 to invalidate the discrimination dictionary JD (or JD ′) and validate the stored command recognition dictionary with the recognized dictionary ID. (See FIGS. 6B and 6C, FIGS. 8B and 8C).

  The voice tag registered in the command recognition dictionary (voice tag recognition dictionary D1) through the above process (FIG. 3) is not shown because it is not related to the gist of the present invention. Can be used to control 5. In this case, in the speech recognition unit 9, the device control signal generation unit 15 outputs a device control signal corresponding to the command recognized by the speech recognition processing unit 12, and performs control corresponding thereto for the control target device. Execute. At that time, control (screen display, change thereof, etc.) corresponding to the command is performed on the display units 25 and 31 displaying the video related to the operation state of the control target device, and the operation of the control target device is performed. Control (change of sound, etc.) corresponding to the command is performed on the speaker 27 (including the wireless headphones 32) that outputs sound related to the state. For example, as illustrated in FIG. 9B, when the user utters “Go to [My Home]” via the microphone 8 while the map screen 67 around the vehicle position is displayed, voice recognition is performed. The unit 9 executes voice recognition processing based on the voice command (PCM data) and outputs a device control signal according to the execution result. In response to the device control signal, the navigation unit 5 displays information such as a guide route to the home (My Home) on the screen of the display unit 25.

  As described above, according to the in-vehicle speech recognition apparatus 10 according to the present embodiment, in registering a command (including a voice tag) input through the microphone 8 in the recognition dictionary D1 and the discrimination dictionary JD, If only the discrimination dictionary JD is used for voice recognition of the command, and the recognition score calculated based on the result is lower than the threshold, the command is similar to any command registered so far. It is determined that there is no (dissimilar) data, and the command is registered in each of the dictionaries D1 and JD.

  On the other hand, if the recognition score is higher than the threshold, it is determined that the command is the same or similar (same / similar) data as any of the registered commands so far, Guidance information (warning screen) 53 notifying registration in each of the dictionaries D1 and JD is displayed.

  That is, it is determined whether the command to be registered (including the voice tag) is the same / similar to or not similar to the command already registered in the discrimination dictionary JD, and the command is registered only when it is dissimilar. If it is the same / similar, it is not registered.

  As a result, it is possible to eliminate the inconvenience such as confusion at the time of recognition caused by registration of voice data that is the same as or similar to a command registered in the past as seen in the prior art. As a result, the recognition rate for voice commands registered in the recognition dictionary can be increased.

  In the above-described embodiment, the case where the in-vehicle speech recognition device 10 is incorporated as a part of the in-vehicle A / V / navigation system 40 has been described as an example, but the gist of the present invention (command (voice tag) issued by the user is a command) When registering in the recognition dictionary (voice tag recognition dictionary D1), a determination dictionary for determining whether the registration should be performed is created, and a recognition process for a command input using the determination dictionary As will be clear from the results (recognition score), only commands that are not the same or similar are registered in the recognition dictionary). As a result, both the A / V device and the navigation device are necessarily included. Of course, it is not necessary to use it by incorporating it into the system.

  In the above-described embodiment, the HDD 7 is used as a recording medium for storing a command recognition dictionary, a discrimination dictionary, and the like together with map data. Instead, another rewritable recording medium such as a flash memory is used. May be used.

1 is a block diagram showing a configuration of an in-vehicle audio / video (A / V) navigation system incorporating an in-vehicle voice recognition device according to an embodiment of the present invention. It is a block diagram which shows the structure of the vehicle-mounted speech recognition apparatus in FIG. It is a flowchart which shows an example of the process which concerns on registration of the voice tag based on the switching selection of the command recognition dictionary and discrimination dictionary performed in the vehicle-mounted speech recognition apparatus of FIG. FIG. 4 is a supplementary explanatory diagram (part 1) of the processing flow of FIG. 3; FIG. 4 is a supplementary explanatory diagram (2) of the processing flow of FIG. 3. FIG. 4 is a supplementary explanatory diagram (part 3) of the processing flow of FIG. 3; FIG. 4 is a supplementary explanatory diagram (4) of the processing flow of FIG. 3. FIG. 6 is a supplementary explanatory diagram (No. 5) of the processing flow of FIG. 3. It is a figure for demonstrating the usage example of a voice tag.

Explanation of symbols

1 to 5 ... controlled devices,
7 HDD (storage means),
8 ... Microphone (voice input means),
9: Voice recognition unit (voice recognition means),
10 ... Vehicle speech recognition device,
12 ... voice recognition processing unit,
13 ... dictionary switching selection unit,
14 ... voice reproduction processing unit,
15 ... Device control signal generator,
20, 30 ... operation unit,
25, 31 ... display unit (display means),
27 ... Speaker,
40. Car audio / video (A / V) navigation system,
53. Guidance information (warning screen) not to register voice tag,
D1, D1 '... Voice tag recognition dictionary (command recognition dictionary),
D2, D3 ... Command recognition dictionary,
JD, JD '... Dictionary for discrimination.

Claims (6)

Voice input means for inputting commands spoken by the user in the passenger compartment;
A plurality of recognition dictionaries in which commands that can be selected according to the operation state of the control target device are used for comparison and collation with commands input via the voice input means, and the input commands are A storage means for storing a determination dictionary for determining whether or not to register in a recognition dictionary and adapted to register the same command as that registered in the plurality of recognition dictionaries;
Dictionary switching selection means for switching the recognition dictionary effective according to the operation state of the control target device;
Voice recognition means for performing voice recognition based on a comparison collation between a command input via the voice input means and a command registered in any dictionary stored in the storage means;
When the command is input via the voice input unit, the voice recognition unit performs voice recognition on the input command using only the discrimination dictionary, and the recognition calculated based on the voice recognition A vehicle-mounted speech recognition apparatus, wherein when the score is lower than a predetermined threshold, the command is registered in the recognition dictionary and the discrimination dictionary selected according to the operation state of the control target device. Furthermore, a display means is provided,
When the recognition score calculated based on the voice recognition is higher than a predetermined threshold, the voice recognition unit warns the display unit that the command is not registered in the recognition dictionary and the discrimination dictionary. The on-vehicle speech recognition apparatus according to claim 1, wherein a screen is displayed.   When the command is input through the voice input unit and the command is not yet registered in the determination dictionary, the voice recognition unit directly uses the input command for the recognition dictionary and the determination. The in-vehicle speech recognition device according to claim 1, wherein the in-vehicle speech recognition device is registered in a dictionary for a vehicle.   The in-vehicle speech recognition according to claim 1, wherein a command supported as a product by a navigation unit cooperating with the in-vehicle speech recognition device is registered in the discrimination dictionary as a reserved word. apparatus. The voice input means for inputting a command spoken by the user in the passenger compartment and the command input via the voice input means are used for comparison and collation, and can be selected according to the operating state of the control target device. A plurality of recognition dictionaries in which commands are registered, and a discrimination dictionary for determining whether or not an input command should be registered in the recognition dictionary, and the same commands as those registered in the plurality of recognition dictionaries are registered Storage means storing the one adapted to be performed, dictionary switching selection means for switching the recognition dictionary effective according to the operating state of the control target device, and a command input via the voice input means, automotive speech and a speech recognition means for performing speech recognition based on the comparison and collation with the commands that are registered in any dictionary that is stored in the storage means In the identification apparatus,
When a command is input via the voice input means, only the discrimination dictionary is enabled and voice recognition is performed on the input command.
When the recognition score calculated based on the voice recognition is lower than a predetermined threshold, the command is registered in the recognition dictionary and the determination dictionary selected according to the operation state of the control target device. Characteristic voice command registration method.   When the recognition score calculated based on the voice recognition is higher than a predetermined threshold value, the display unit displays a warning screen indicating that the command is not registered in the recognition dictionary and the determination dictionary. The voice command registration method according to claim 5.

Images