JP4770374B2 - Voice recognition device - Google Patents

Description

  The present invention relates to a voice recognition device, and more particularly to a voice recognition device that recognizes voice commands for a plurality of control target devices.

The speech recognition apparatus improves the recognition rate by limiting the recognition vocabulary according to the operation state of the control target device. That is, a plurality of recognition dictionaries having a predetermined recognition vocabulary corresponding to the operation state of the control target device are stored, the operation state of the control target device is detected, and a voice recognition process is performed from the stored plurality of recognition dictionaries. The recognition dictionary used for is selected.
Japanese Patent Laid-Open No. 10-7348

  When the voice recognition device recognizes voice commands for a plurality of devices to be controlled, a plurality of recognition dictionaries used for voice recognition processing may be selected. For example, in the case of a voice recognition device for a vehicle, when a navigation device and an audio device are included as devices to be controlled and both devices are operating, a recognition dictionary based on the operation state of the navigation device ( For example, a recognition dictionary for destination setting operation or a recognition dictionary for search operation) and a recognition dictionary based on the operating state of the audio device (for example, a recognition dictionary for music performance operation) are selected as the dictionary used for the speech recognition processing. The

  In the recognition vocabulary of the plurality of recognition dictionaries, there are common words included in common in the plurality of recognition dictionaries. Examples of the common language include “previous” and “next”. The recognition dictionary for destination setting operation includes the recognition word “previous (next) destination”, which is recognized for music performance operation. The dictionary contains the recognition word “previous (next) song”.

  When a plurality of dictionaries are selected as recognition dictionaries for use in the speech recognition processing, and common words are included in the selected dictionaries, other words including the common words may be mistakenly recognized. It was. For example, the user may be recognized as the “previous song” even though the user has said “the previous destination”. Also, if you say “Previous” without thinking of the voice command “Previous destination”, the voice command you wanted to say is not recognized, and other commands such as “Previous song” Sometimes it was recognized.

  In addition, it is possible to have all recognition words for all control target devices in one recognition dictionary without providing a plurality of recognition dictionaries, but in that case, naturally, if the user's utterance includes a common word, The aforementioned misrecognition problem arises.

  The present invention has been made based on this situation, and an object of the present invention is to provide a speech recognition apparatus with high recognition accuracy of speech commands including common words.

In order to achieve the object, the invention according to claim 1 is a voice recognition device for recognizing a voice command for a plurality of control target devices, wherein a plurality of recognition words are respectively among the plurality of control target devices. Based on the storage device storing the recognition dictionary associated with the operation state of any one of the control target devices, and the recognition dictionary stored in the storage device, the voice command spoken by the user is recognized It is determined whether or not a common word that is included in common in at least two recognition words associated with the voice command recognition unit and the different control target devices is included in the voice command spoken by the user. Common language recognition means
The voice command recognizing means, when the common word recognizing means determines that the common word is included , the control last operated by the user among the recognized words of the recognition dictionary including the common word. A voice command uttered by the user is recognized based on a recognition word associated with the target device.

According to the first aspect of the present invention, it is determined whether or not a common word is included in a voice command uttered by the user by the common word recognition unit. If it is determined that the word is included , the user speaks based on the recognition word associated with the device to be controlled last operated by the user among the recognition words of the recognition dictionary including the common word. The voice command is recognized. Therefore, when it is determined that a common word is included in the user's utterance, a recognized word that is not associated with the device to be controlled last operated by the user is not recognized. The recognition accuracy of the voice command is improved.

  The storage device may store only one recognition dictionary having all the recognition words for all the control target devices. However, in order to improve the recognition rate, the recognition dictionary is classified according to the operation state of the control target device. It is preferable to be divided into a plurality of recognition dictionaries. In that case, the recognition accuracy of the voice command including the common word is improved as described in claim 2.

  That is, the invention according to claim 2 is the speech recognition apparatus according to claim 1, wherein the recognition dictionary stored in the storage device is divided into a plurality of recognition dictionaries according to operation states of the control target devices. , Further comprising dictionary selection means for selecting at least one recognition dictionary from the plurality of recognition dictionaries stored in the storage device based on the operating state of the plurality of control target devices, When it is determined by the common word recognition means that the common word is included, as the recognition dictionary for recognizing the voice command spoken by the user, the user is the last among the recognition dictionaries selected by the dictionary selection means. One recognition dictionary determined based on the operated device to be controlled is used.

  According to the second aspect of the present invention, it is determined whether or not a common word is included in a voice command uttered by the user by the common word recognition unit. When it is determined that it is included, one recognition dictionary determined based on the control target device last operated by the user is used as a recognition dictionary for recognizing a voice command. For this reason, even if a plurality of recognition dictionaries are selected from the operating state of the control target device, only one recognition dictionary determined based on the control target device last operated by the user is used by the voice command recognition unit. As a result, the recognition accuracy of voice commands including common words is improved.

  The voice command recognizing means can determine the voice command spoken by the user as described in claim 3 or 4.

  According to a third aspect of the present invention, in the voice recognition device according to the second aspect, when the voice command recognition unit narrows down the recognition dictionary to one based on the device to be controlled last operated by the user, Among the recognition words included in the narrowed-down recognition dictionary, the recognition word including the common word recognized by the common word recognition means is specified as the voice command uttered by the user.

  According to a fourth aspect of the present invention, in the voice recognition device according to the second aspect, the voice command recognition means narrows down the recognition dictionary to one based on the device to be controlled last operated by the user. Is for recognizing a voice command uttered by a user by executing a voice recognition process using the narrowed-down recognition dictionary.

  Further, as described in claim 3, when the common word recognized by the common word recognition means and the recognition vocabulary of the recognition dictionary are compared, the voice command spoken by the user is specified. As described in 5, the common word recognition means is preferably executed regardless of the number of recognition dictionaries selected by the dictionary selection means.

  The user may speak only a common word regardless of the operation state of the control target device. However, according to the fifth aspect, only one recognition dictionary is determined from the operation state of the control target device. Even if it is selected, the common word recognition means is executed. By executing the common word recognizing means, even when the user utters only the common word, the common word can be recognized with high accuracy, and the recognition vocabulary of the common word and the recognition dictionary can be recognized with high accuracy. Therefore, the voice command is specified, so that the accuracy of specifying the voice command is improved.

  Further, as in claim 6, when a plurality of recognition dictionaries are selected by the dictionary selecting means, and at least two of the selected recognition dictionaries include the common word In addition, the common word recognition means may be executed.

  According to the sixth aspect, the common word recognizing unit is not executed depending on the condition, and the voice command spoken by the user is directly recognized by the voice command recognizing unit, so that the processing becomes faster.

  Further, the common word recognizing means includes a common word recognition dictionary as described in claim 7, or uses a word spotting technique as described in claim 8 to share a voice command spoken by a user. Determine if a word is included.

  That is, the invention according to claim 7 is the speech recognition device according to any one of claims 1 to 6, wherein a common word recognition dictionary having the common word as a recognition vocabulary is stored in the storage device. The common word recognition means determines whether or not the common word is included in a voice command uttered by a user by executing a voice recognition process using the common word recognition dictionary.

  According to an eighth aspect of the present invention, in the speech recognition apparatus according to any one of the first to sixth aspects, the common word recognition means includes an input voice representing a user's utterance and a reference for the common word stored in advance. By comparing with voice, it is determined whether or not a common word is included in the voice command spoken by the user.

  In the first to eighth aspects of the invention, first, after determining whether or not a common word is included in the voice command spoken by the user, the voice command is recognized based on the recognition dictionary. As described in claim 9, after selecting a voice command candidate using a recognition dictionary, it may be determined whether or not a common word is included in the selected candidate.

That is, the invention according to claim 9 is a voice recognition device for recognizing a voice command for a plurality of control target devices, wherein a plurality of recognition words are controlled by any one of the plurality of control target devices. A storage device that stores a recognition dictionary associated with the operation state of the target device, and a voice command recognition unit that recognizes a voice command spoken by the user based on the recognition dictionary stored in the storage device; A list storage device that stores a list of common words that are commonly included in at least two recognition words associated with different control target devices,
The voice command recognition means includes candidate selection means for selecting a voice command candidate spoken by the user from the recognition dictionary by voice recognition processing using the recognition dictionary, and a plurality of candidates are selected by the candidate selection means. And when at least two of the selected candidates include a common word included in the common word list, the common word included in the common word list is a candidate containing, by selecting a candidate user is a recognition vocabulary of the recognition dictionary determined based on the control target device has been operated last, and characterized in that for identifying the speech uttered by a user To do.

Thus, first, when the candidate selection means selects a voice command candidate by voice recognition processing using a recognition dictionary, and a common word is included in at least two of the selected candidates. Even if the user selects a candidate that includes a common word included in the common word list and is a recognition vocabulary of a recognition dictionary that is determined based on the device to be controlled last, Since the voice command related to the device to be controlled last operated is not selected, the recognition accuracy of the voice command including the common word is improved.

  Also in the case of the ninth aspect of the invention, in order to improve the recognition rate, it is preferable that the recognition dictionary is divided into a plurality of recognition dictionaries according to the operation state of the control target device.

  The invention according to claim 10 is the speech recognition apparatus according to claim 9, wherein the recognition dictionary stored in the storage device is divided into a plurality of recognition dictionaries according to operation states of the control target devices, The apparatus further comprises dictionary selection means for selecting at least one recognition dictionary from a plurality of recognition dictionaries stored in the storage device based on operating states of a plurality of control target devices, and the candidate selection means is adapted for voice recognition processing. The recognition dictionary selected by the dictionary selection means is used.

  Preferably, as described in claim 11, a common word recognition dictionary having the common word as a recognition vocabulary is stored in the storage device, and the candidate selecting unit is selected by the dictionary selecting unit. Using the common word recognition dictionary in addition to the recognition dictionary, a voice command candidate spoken by the user is selected.

  In this way, if the candidate selection means further uses the common word recognition dictionary, it becomes possible to accurately recognize that the user has spoken a voice command including the common word, and as a result, the voice including the common word. Instruction recognition accuracy is further improved.

  Embodiments of the apparatus of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an in-vehicle navigation device 10 having a function as a voice recognition device of the present invention.

  The on-vehicle navigation device 10 includes a microphone 12, a speaker 14, an external device control unit 16, a display device 18, and an ECU 100.

  This in-vehicle navigation device 10 has a function of controlling external devices such as an audio, an air conditioner, and a telephone in addition to the function as a navigation device, and the external device control unit 16 is for controlling these external devices. is there.

  The display device 18 is constituted by, for example, a liquid crystal display, and is disposed at an intermediate position between the driver seat and the passenger seat of the instrument panel. In addition to the configuration shown in FIG. 1, the in-vehicle navigation device 10 includes a configuration that a normal navigation device has, such as an operation switch group, a wireless communication device that performs wireless communication with the outside of the vehicle, a DVD A map data input device for inputting map data from a storage medium such as a ROM, a position detector for detecting the current position of the vehicle by a geomagnetic sensor or a gyroscope, a remote controller, a remote control sensor, and the like are provided. Signals from these devices are input to the ECU 100. The ECU 100 also receives signals from various sensors mounted on the vehicle, such as a vehicle speed sensor and a temperature sensor.

  The ECU 100 is a computer including a CPU, a ROM, a RAM, and the like. The ECU 100 recognizes a sound input from the microphone 12, outputs a predetermined sound (including sound) from the speaker 14, or outputs a signal input from the outside. Based on this, the control content of the external device is determined, and the determined control content is output to the external device control unit 16, or the display screen of the display device 18 is controlled based on a signal input from the outside.

  Next, the internal configuration of the ECU 100 will be described. The voice output control unit 102 receives the user's voice detected by the microphone 12. When the signal is input from the microphone 12, the voice output control unit 102 utters the user's speech based on the input signal. Check. When the utterance can be confirmed, the sound from the microphone 12 is output to the language analysis processing unit 104. The audio output control unit 102 can exchange signals with the navigation interface unit 106, and the output determined by the destination search unit 108, the guidance / search control unit 110, and the memory point control unit 112. A sound command signal is supplied via the navigation interface unit 106. When the output sound command signal is supplied, a predetermined output sound signal is generated based on the signal and is output to the speaker 14, thereby outputting a predetermined sound from the speaker 14.

 The language analysis processing unit 104 analyzes the generated content confirmed by the voice output control unit 102 and inputs it to the voice recognition engine 114. The voice recognition operation control unit 116 includes a storage device 117, and the storage device 117 stores a plurality of recognition dictionaries corresponding to the operation state of the control target device. The device to be controlled means the in-vehicle navigation device 10 or the external device described above.

  Examples of the plurality of recognition dictionaries include a recognition dictionary corresponding to the destination setting mode of the in-vehicle navigation device 10, a recognition dictionary corresponding to the audio song selection mode, a recognition dictionary corresponding to the air conditioner setting operation mode, and the speaker 14 There is a recognition dictionary corresponding to the sound mode (which indicates which of the devices that can output sound, such as FM radio, AM radio, CD player, etc.). In these recognition dictionaries, for example, the recognition dictionaries corresponding to the purpose setting mode include recognition words such as “previous destination” and “next destination” in the recognition vocabulary, and an audio song selection mode. In the recognition dictionary corresponding to, “next song” and “previous song” are included in the recognition vocabulary. Thus, “next”, “previous”, and the like are words that are commonly included in a plurality of recognition dictionaries, that is, common words. The storage device 117 also stores a common word recognition dictionary that uses this common word as a recognition vocabulary. The storage device 117 functions as a list storage device, and a common language list is also stored.

  The voice recognition operation control unit 116 receives signals from the destination search unit 108, the guidance / search control unit 110, the memory point control unit 112, the external device control unit 16, and the display device 18 via the navigation interface unit 106. Is done. The last management information stored in the last mode management unit 120 is also input to the voice recognition operation control unit 116 via the navigation interface unit 106. The last management information means information that manages (stores) what the control target device last operated by the user was.

  The voice recognition operation unit 116 determines the currently operated control target device and the operation state of the currently operated control target device based on the input signals, and stores them in the storage device 117 based on the determined operation target device. The speech recognition engine 114 selects a recognition dictionary to be used for speech recognition processing from the plurality of recognition dictionaries.

  The voice recognition engine 114 recognizes (determines) which voice command is the generated content analyzed by the language analysis processing unit 104 using the recognition dictionary selected by the voice recognition operation control unit 116.

  The communication control unit 120 notifies the external device control unit 16 of a voice command recognized by the voice recognition engine 114 and a control command input by operating an operation switch group (or a remote controller) (not shown). The operation state of the external device input from the external device control unit 16 is input to the navigation interface control unit 106.

  The display content determined by the destination search unit 108, the guidance / search control unit 110, and the memory point control unit 112 is supplied to the display control unit 122 via the navigation interface unit 106, and displayed according to the supplied content. The display screen of the device 18 is controlled. The display device 18 also displays a recognition result in the speech recognition engine 114. Further, the display device 18 can exchange signals with the external device control unit 16, and the display screen of the display device 18 displays the state of the external device based on the signal from the external device control unit 16. Is also displayed.

  FIG. 2 is a flowchart showing a process for recognizing a voice command uttered by the user among the process contents in the ECU 100. The process shown in FIG. 2 is executed when a signal is input from the microphone 12 and the speech output control unit 102 confirms the user's utterance.

  In FIG. 2, first, in step S10, it is determined whether or not the recognition level is appropriate. This determination is denied when the input sound from the microphone 12 is too loud or too small. On the other hand, if the determination in step S10 is affirmative, one or more of the plurality of recognition dictionaries stored in the storage device 117 is selected based on the operation state of the control target device in step S20 corresponding to the dictionary selection unit. Select multiple recognition dictionaries.

  In subsequent step S30, it is determined whether or not a plurality of recognition dictionaries are selected in step S20, and whether or not the same common word is included in at least two of the plurality of recognition dictionaries. To do.

  If step S30 is affirmed, it is further determined in step S40 whether or not there is last management information, that is, whether or not the last management information is stored in the last mode management unit 118. If this determination is also affirmed, speech recognition processing is executed using the common word recognition dictionary stored in the storage device 117 in step S50 corresponding to the common word recognition means. This speech recognition process calculates likelihood by comparing speech data of a recognized word included in a recognition dictionary as a recognition vocabulary with a user's uttered speech, and a recognized word having a predetermined likelihood or higher. If there is (a common language here), it is determined that the recognized word is a spoken word. When there are a plurality of recognized words having a predetermined likelihood or more, the recognized word having the highest likelihood is determined as the spoken word.

  In the subsequent step S60, it is determined whether or not it has been recognized that there is a common word as a result of executing step S50. If the determination in step S60 is negative, and if the determinations in steps S30 and S40 described above are negative, in step S70, a voice command recognition process similar to the conventional one is executed. That is, the speech recognition process is executed using one or a plurality of recognition dictionaries selected based on the operation state of the control target device. Thereby, one voice command is determined as the voice command spoken by the user.

  On the other hand, if the determination in step S60 is affirmative, step S80 is executed. In step S80, the recognition dictionary is narrowed down based on the last management information stored in the last mode management unit 118. That is, in step S80, one recognition dictionary is selected from the recognition dictionary selected in step S20 using the last management information stored in the last mode management unit 118.

  In subsequent step S90, the second speech recognition process is executed using the one recognition dictionary selected in step S80. This determines what the voice command the user has spoken. In FIG. 2, steps S40 and S60 to 90 correspond to voice command recognition means.

  In the speech recognition process in step S90, as in the process in step S50, the user utters using the likelihood determined based on the comparison between the speech data of the recognized word contained in the recognition dictionary and the user's speech. Determine what the voice command is. Therefore, even if the recognized word included in the recognition dictionary and the word spoken by the user do not completely match, if the likelihood is equal to or greater than the predetermined value, the voice command spoken by the user is determined. For example, even when the word spoken by the user is simply “previous”, “previous destination” is determined as the word spoken by the user depending on the likelihood.

  After executing step S70 or step S90 and determining the voice command spoken by the user, in step S100, it is determined whether or not the process is successful. This determination is to determine whether or not the voice command determined in step S70 or S90 can be executed. If the voice command can be executed, step S100 is affirmed and the voice command cannot be executed. Step S100 is denied. For example, in the state where the route to the destination has not been determined, if an instruction to change the route is recognized, the determination in step S100 is denied.

  If the determination in step S100 is affirmed, in a subsequent step S110, a function based on the voice command determined in step S70 or S90 is executed by a predetermined control target device.

  On the other hand, if the determination in step S100 is negative, a predetermined error notification is output from the display device 18 and the speaker 14 in step S120. For example, if step S10 is denied, a message indicating that the voice command cannot be recognized because the recognition level is inappropriate is output. If step S100 is denied, the message is input in the current operating state. A message indicating that the voice command cannot be executed is output.

  As described above, according to the present embodiment described above, it is determined whether or not a common word is included in the voice command spoken by the user by executing the common word recognition process in step S50. If it is determined that the recognition dictionary is determined, the recognition dictionary is narrowed down to one based on the device to be controlled last operated by the user (step S80), and voice recognition processing is performed using the narrowed recognition dictionary (step S80). Step S90). Therefore, for example, even if two (or more) recognition dictionaries are selected in step S20 because an audio operation is being performed and a destination setting operation is being performed, in step S50, “next” “ If a common word such as “previous” is recognized, the recognition dictionary is narrowed down to one based on the last management information in step S80, and then the speech recognition process is executed in step S90. As a result, for example, when the user's last operating device is audio, it is prevented that the “previous destination” is recognized even though the user has said “previous song”. Of course, even if you simply say “previous” without thinking of the command “previous song”, the “previous destination” not intended by the user will not be recognized, Recognition accuracy of voice commands including common words is improved.

  Next, a second embodiment of the present invention will be described. The second embodiment is an in-vehicle navigation device having a configuration similar to that of FIG. 1, but the processing of the ECU 100 for specifying a voice command spoken by the user is different from that of the first embodiment. FIG. 3 is a flowchart showing processing for specifying a voice command spoken by the user, which is executed by the ECU 100 in the second embodiment.

  3 is different from that shown in FIG. 2 described above in that step S30 is not executed but step S20 is directly executed, and then step S40 is directly executed, instead of step S50 in FIG. Step S55 is executed, and step S95 is executed instead of step S90 in FIG. Hereinafter, this difference will be described.

  In the second embodiment, after selecting one or more recognition dictionaries from the plurality of recognition dictionaries stored in the storage device 117 based on the operation state of the control target device in step S20, the number of selected dictionaries is Step S40 is executed without determining whether or not there are a plurality. Therefore, even if only one dictionary is selected, step S40 and subsequent steps are executed.

  If it is determined in step S40 that there is last management information, step S55 corresponding to the common word recognition unit is executed. In step S55, the user speaks using a known word spotting method such as a method using DP matching based on the speech waveform input from the microphone 12 and the speech waveform of the common word stored in advance. Attempts to extract common words from voice commands. If the common word can be extracted in step S55, the following step S60 is affirmed, and if the common word cannot be extracted, step S60 is denied.

  In step S80, the recognition dictionary is narrowed down based on the last management information stored in the last mode management unit 118, as in the case of the first embodiment. In the second embodiment, step S80 is executed even when only one recognition dictionary is selected in step S20. However, if there is already only one recognition dictionary, in step S80, That one recognition dictionary is selected.

  In the subsequent step S95, the common word extracted in step S55 is compared with the recognition vocabulary of the recognition dictionary narrowed down in step S80, and the common word extracted in step S55 from the recognition vocabulary of the recognition dictionary narrowed down in step S80. Is identified as a voice command spoken by the user. After this step S95 is executed, the above-described step S100 and subsequent steps are executed.

  As described above, in the case of the second embodiment described above, even if only one recognition dictionary is selected in step S20, common words are extracted from the utterance contents of the user by executing S55. Therefore, even when the user utters only a common word, it can be recognized with high accuracy. In step S95, the common word that can be recognized with high accuracy is compared with the recognition vocabulary of the recognition dictionary. Since the voice command is specified, the accuracy of specifying the voice command including the common word is improved.

  Next, a third embodiment of the present invention will be described. The third embodiment is also an in-vehicle navigation device having the same configuration as that in FIG. 1, but the processing of the ECU 100 for specifying the voice command spoken by the user is different from the first and second embodiments. FIG. 4 is a flowchart showing processing for recognizing a voice command spoken by the user, which is executed by the ECU 100 in the third embodiment.

  In FIG. 4, first, the same steps S10 and S20 as in FIG. 2 are executed. After one or more recognition dictionaries are selected in step S20, voice command candidates are selected using the recognition dictionaries selected in step S20 in step S200 corresponding to the candidate selection means. That is, the speech recognition process using the recognition dictionary selected in step S20 is executed, and among the recognition words included in the recognition vocabulary of the recognition dictionary as a result of the processing, a recognition word having a likelihood equal to or higher than a predetermined value is selected as a voice command. Select as a candidate.

  In subsequent step S210, by comparing the candidate selected in step S200 with the list of common words stored in the storage device 117, is the same common word included in at least two of the selected candidates? Judge whether or not.

  If the determination in step S210 is affirmative, step S220 is executed. In step S220, it is further determined whether or not there is last management information, that is, whether or not the last management information is stored in the last mode management unit 118. If the determination in step S220 is negative and if the determination in step S210 is negative, step S230 is executed. In step S230, it is determined that the highest likelihood among the plurality of candidates selected in step S210 is the voice command spoken by the user.

  On the other hand, if the determination in step S220 is affirmative, the candidate includes a common word among the plurality of candidates selected in step S210, and is a recognition vocabulary of a recognition dictionary determined based on the last management information. The candidate is determined to be a voice command spoken by the user. For example, in step S210, a plurality of candidates including the common word “next” such as “next song” and “next destination” are selected, and a recognition dictionary determined based on the last management information is a recognition dictionary for destination setting. In some cases, the voice command spoken by the user is determined to be the “next destination”. In FIG. 4, steps S200 to S240 correspond to voice command recognition means.

  After identifying what the voice command the user uttered in step S230 or S240, the same steps as in the first embodiment are executed after step S100.

  As described above, as in the third embodiment described above, first, a speech command candidate is selected by speech recognition processing using a recognition dictionary (S200), and then common to at least two of the selected candidates. It is determined whether or not a word is included (S210), and when a common word is included, one candidate is selected based on the device to be controlled last operated (S240). Is not selected except for the voice command related to the device to be controlled last operated, so that the recognition accuracy of the voice command including the common word is improved.

  Since the common word recognition dictionary is stored as a recognition dictionary in the storage device 117, when the user utters a voice command including the common word, the candidate selected in step S200 is a candidate for only the common word. Will also be included. As a result, when the user utters a voice command including a common word, step S210 is more reliably affirmed, and as a result, the recognition accuracy of the voice command including the common word is further improved.

  As mentioned above, although embodiment of this invention was described, this invention is not limited to the above-mentioned embodiment, The following embodiment is also contained in the technical scope of this invention, and also the summary other than the following is also included. Various modifications can be made without departing from the scope.

  For example, in the above-described embodiment, a plurality of recognition dictionaries are stored for each operation state of the control target device. Instead of associating each recognition dictionary with the operation state of the control target device, each recognition word included in the recognition dictionary is changed. The recognition dictionaries may be combined into one by associating with the operation state of the control target device.

It is a block diagram which shows the structure of the vehicle-mounted navigation apparatus 10 provided with the function as a voice recognition apparatus of this invention. It is a flowchart which shows the process for recognizing the voice command which the user uttered among the processing content in ECU100 of FIG. It is a flowchart which shows the process for recognizing the voice command which the user uttered which ECU100 performs in 2nd Embodiment. It is a flowchart which shows the process for recognizing the voice command which the user uttered which ECU100 performs in 3rd Embodiment.

Explanation of symbols

10: In-vehicle navigation device (voice recognition device)
100: ECU
117: Storage device S20: Dictionary selection means S50: Common word recognition means S55: Common word recognition means S40, S60 to S90: Voice command recognition means S200: Candidate selection means S200 to S240: Voice command recognition means

Claims (11)

A voice recognition device for recognizing voice commands for a plurality of control target devices,
A storage device storing a recognition dictionary in which a plurality of recognition words are respectively associated with operation states of any one of the plurality of control target devices;
Voice command recognition means for recognizing a voice command spoken by the user based on a recognition dictionary stored in the storage device;
Common word recognition means for determining whether or not a common word included in common in at least two recognition words associated with different control target devices is included in a voice command spoken by a user; With
The voice command recognizing means, when the common word recognizing means determines that the common word is included , the control last operated by the user among the recognized words of the recognition dictionary including the common word. A voice recognition device, wherein a voice command spoken by a user is recognized based on a recognition word associated with a target device. The recognition dictionaries stored in the storage device are divided into a plurality of recognition dictionaries according to the operating state of the device to be controlled,
Further comprising a dictionary selection means for selecting at least one recognition dictionary from a plurality of recognition dictionaries stored in the storage device based on an operation state of the plurality of control target devices;
The voice command recognizing means, when the common word recognizing means determines that a common word is included, the recognition selected by the dictionary selecting means as a recognition dictionary for recognizing a voice command spoken by the user The speech recognition apparatus according to claim 1, wherein one recognition dictionary determined based on a device to be controlled last operated by a user is used.   When the voice command recognition means narrows down the recognition dictionary to one based on the device to be controlled last operated by the user, the common word recognition among the recognition words included in the narrowed one recognition dictionary The speech recognition apparatus according to claim 2, wherein a recognition word including a common word recognized by the means is specified as a voice command spoken by a user.   When the voice command recognizing means narrows down the recognition dictionary to one based on the device to be controlled last operated by the user, by executing the voice recognition process using the narrowed down recognition dictionary, The voice recognition apparatus according to claim 2, which recognizes a voice command spoken by a user.   The speech recognition apparatus according to claim 3, wherein the common word recognition unit is executed regardless of the number of recognition dictionaries selected by the dictionary selection unit.   The common word recognition unit is executed when a plurality of recognition dictionaries are selected by the dictionary selection unit and the common word is included in at least two of the selected plurality of recognition dictionaries. The speech recognition apparatus according to claim 2, wherein A common word recognition dictionary having the common word as a recognition vocabulary is stored in the storage device;
The common word recognition means determines whether or not the common word is included in a voice command spoken by a user by executing voice recognition processing using the common word recognition dictionary. The voice recognition device according to claim 1.   The common word recognition unit determines whether or not a common word is included in a voice command uttered by the user by comparing an input voice representing the user's utterance with a reference voice of the common word stored in advance. The speech recognition apparatus according to claim 1, wherein A voice recognition device for recognizing voice commands for a plurality of control target devices,
A storage device storing a recognition dictionary in which a plurality of recognition words are respectively associated with operation states of any one of the plurality of control target devices;
Voice command recognition means for recognizing a voice command spoken by the user based on a recognition dictionary stored in the storage device;
A list storage device that stores a list of common words that are commonly included in at least two recognition words that are associated with different control target devices;
The voice command recognition means includes candidate selection means for selecting a voice command candidate spoken by the user from the recognition dictionary by voice recognition processing using the recognition dictionary, and a plurality of candidates are selected by the candidate selection means. And when at least two of the selected candidates include a common word included in the common word list , the common word included in the common word list is a candidate containing, by selecting a candidate user is a recognition vocabulary of the recognition dictionary determined based on the control target device has been operated last, and characterized in that for identifying the speech uttered by a user Voice recognition device. The recognition dictionaries stored in the storage device are divided into a plurality of recognition dictionaries according to the operating state of the device to be controlled,
Further comprising a dictionary selection means for selecting at least one recognition dictionary from a plurality of recognition dictionaries stored in the storage device based on an operation state of the plurality of control target devices;
The speech recognition apparatus according to claim 9, wherein the candidate selection unit uses a recognition dictionary selected by the dictionary selection unit in speech recognition processing. A common word recognition dictionary having the common word as a recognition vocabulary is stored in the storage device;
11. The candidate selection unit is configured to select a voice command candidate spoken by a user using the common word recognition dictionary in addition to the recognition dictionary selected by the dictionary selection unit. The speech recognition apparatus described in 1.

Images