Speech recognition device and speech recognition method

Abstract

A voice recognition device accepts a control request for a device mounted on a mobile body based on an instruction given by voice uttered by a user, and the voice recognition device comprises: a voice acquisition unit that acquires voice; a voice data conversion unit that converts the acquired voice into voice data; an operation object device determination unit that analyzes the voice data to determine the operation object device that is the object of the operation; a detection mode setting unit that sets a detection mode for determining a control request corresponding to the voice data based on the operation object device; and a control request determination unit that analyzes the voice data based on the set detection mode to determine the control request for the operation object device.

Description

Technical Field

[0001] The invention relates to a speech recognition device and a speech recognition method. Background Art

[0002] The speech recognition device recognizes the speech uttered by the user and performs various processing based on the speech recognition result. The vehicle equipment control device described in Patent Document 1 has a keyword detection unit for detecting keywords including ambiguous words, and can perform speech recognition as close to the user's intention as possible even when the instruction given by the speech uttered by the user is unclear.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Publication No. 2006-308848 Summary of the invention

[0006] Problem that the invention aims to solve

[0007] However, when controlling a device mounted on a mobile body such as a vehicle by the vehicle equipment control device described in Patent Document 1, it may be inappropriate to perform control based on unclear instructions depending on the type of the device.

[0008] An object of the present invention is to provide a speech recognition device and a speech recognition method capable of identifying a control request based on an instruction given by a user's speech in an appropriate detection mode corresponding to each operation target device.

[0009] Solutions for solving problems

[0010] The speech recognition device and speech recognition method according to the present invention solve the above-mentioned problems by setting a detection mode for determining a control request according to an operation target device and determining the control request based on the set detection mode.

[0011] Effects of the Invention

[0012] According to the present invention, since the detection mode can be set according to the operation target device, there is an effect of determining a control request based on an instruction by a user's voice in an appropriate detection mode corresponding to each operation target device. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 This is a block diagram showing the configuration of a vehicle using the voice recognition device according to the first embodiment of the present invention.

[0014] Figure 2 It means use Figure 1The flowchart of the speech recognition method of the speech recognition device is shown.

[0015] Figure 3 It means by Figure 1 The diagram shows an example of a first detection mode target device and a second detection mode target device which are preset in a speech recognition device.

[0016] Figure 4 is a table illustrating the difference in speech recognition range between the second detection mode based on rule base recognition and the first detection mode based on natural language recognition.

[0017] Figure 5 It is a block diagram showing the configuration of a vehicle including a voice recognition device according to a second embodiment of the present invention.

[0018] Figure 6 It is a diagram showing the configuration of a vehicle using a voice recognition device according to a third embodiment of the present invention.

[0019] Figure 7 It means use Figure 6 The flowchart of the speech recognition method of the speech recognition device is shown.

[0020] Figure 8 It means in Figure 7 A table showing the scores of the misrecognition levels added according to the respective vehicle states in the voice recognition method shown.

[0021] Fig. 9 It means in Figure 7 A diagram showing the correspondence between the operation target device and the detection mode when it is determined that the voice collection environment is not good in the voice recognition method shown.

[0022] Fig.10 It means use Figure 6 A flowchart of another example of a speech recognition method of the speech recognition device shown. DETAILED DESCRIPTION

[0023] Hereinafter, embodiments of the present invention will be described based on the drawings.

[0024] <<First Implementation Method>>

[0025] Reference Figures 1 to 4 The speech recognition device 100 according to the first embodiment and the speech recognition method using the speech recognition device 100 will be described.

[0026] like Figure 1As shown, a vehicle 1 as a mobile body has an interior lamp 3, an air conditioner 4, a door and window control device 5, a sunroof control device 6, a back door control device 7, a display 12, an audio device 13, a navigation device 14, a communication device 15, a travel control device 31, a headlight 32, a seat control device 33, and a trunk control device 34 as vehicle-mounted devices. In addition, the vehicle 1 has a device control unit 11 for controlling these vehicle-mounted devices, a voice acquisition unit 101 as a microphone for inputting instructions by voice of a user, and a communication unit 107 connected to the device control unit 11 and the voice acquisition unit 101 by wireless or wired connection and capable of communicating with a server 20 outside the vehicle.

[0027] In addition, the interior lamp 3, the air conditioner 4 and the audio device 13 are environment adjustment devices for adjusting the environment state inside the vehicle 1, that is, the room. That is, the interior lamp 3 adjusts the environment state including the light quantity. The air conditioner 4 adjusts the environment state including the temperature. The audio device 13 adjusts the environment state including the volume.

[0028] In addition, the communication unit 107 of the vehicle 1 can communicate with the server 20 by wireless. In addition, a server-side communication unit (not shown) capable of communicating with the communication unit 107 of the vehicle 1 is also provided in the server 20. The server 20 has a voice data conversion unit 102, an operation object device determination unit 104, a detection mode setting unit 105, and a control request determination unit 106. The display 12, the audio device 13, the navigation device 14, the call device 15, the device control unit 11, the voice acquisition unit 101, and the communication unit 107 provided in the vehicle 1 and the voice data conversion unit 102, the operation object device determination unit 104, the detection mode setting unit 105, and the control request determination unit 106 of the server 20 constitute a navigation system 10. The navigation system 10 is, for example, an IVI system. In addition, the voice acquisition unit 101 and the communication unit 107 provided in the vehicle 1 and the voice data conversion unit 102, the operation object device determination unit 104, the detection mode setting unit 105 and the control request determination unit 106 of the server 20 constitute a voice recognition device 100. That is, the voice recognition device 100 constitutes a part of the navigation system 10. In addition, the voice recognition device 100 accepts instructions made by voice uttered by the user, and controls any one or more or all of the vehicle-mounted devices via the device control unit 11. That is, the navigation system 10 has a voice recognition device 100, and the voice recognition device 100 accepts control requests for various devices mounted on the vehicle 1 based on instructions made by voice uttered by the user riding in the vehicle 1. The voice recognition device 100 is a voice recognition system.

[0029] In addition, if Figure 3As shown, the door window control device 5 has a door window locking device 5a and a door window opening and closing device 5b. The sunroof control device 6 has a sunroof locking device 6a and a sunroof opening and closing device 6b. The back door control device 7 has a back door locking device 7a and a back door opening and closing device 7b. The trunk control device 34 has a trunk locking device 34a and a trunk opening and closing device 34b.

[0030] The seat control device 33 is a device for controlling various functions of a seat provided in the interior of the vehicle 1, and includes a seat massage device 33a, a seat heater 33b, and a seat position adjustment device 33c. The seat position adjustment device 33c is capable of adjusting the seat height, the seat front-rear position, and the seat back angle.

[0031] in addition, Figure 1 The travel control device 31 shown is a device that controls the travel of the vehicle 1 when the automatic driving mode is selected as the travel mode of the vehicle 1. Specifically, the travel control device 31 controls the driving actuator, the steering actuator, and the brake actuator of the vehicle 1.

[0032] The display 12 is a touch panel display with an input function that can input control instructions for the audio device 13, the navigation device 14, the communication device 15, and other in-vehicle devices through a touch operation of the user. In addition, the display 12 can output information about incoming calls to the communication device 15, information about the other party in a conversation via the communication device 15, and guidance information from the navigation device 14 in the form of images including character information and icons. In addition, the audio device 13 can output voice for notifying the communication device 15 of an incoming call, the voice of the other party in a conversation via the communication device 15, and voice indicating guidance information from the navigation device.

[0033] In addition, the display 12 is not limited to a touch panel display with an input function, as long as it is a device that outputs a 2D image or 3D image including character information. In addition, the display 12 is not limited to a display provided on the dashboard of the vehicle 1, and can be a head-up display projected on the front windshield, or other various displays.

[0034] The display 12 and the audio device 13 constitute information providing means.

[0035] The voice data conversion unit 102 of the server 20 receives the voice information acquired by the voice acquisition unit 101 via the communication unit 107, and converts the voice information into voice data. The voice data is text data based on the voice.

[0036] In addition, the operation object device determination unit 104 of the server 20 analyzes the voice data converted by the voice data conversion unit 102, and determines the operation object device to be operated based on the keywords included in the voice data. The operation object device is, for example, one or more vehicle-mounted devices among the interior lamp 3, the air conditioner 4, the door and window control device 5, the sunroof control device 6, the back door control device 7, the display 12, the audio device 13, the navigation device 14, the communication device 15, the driving control device 31, the headlight 32, the seat control device 33, and the trunk control device 34. In addition, the operation object device determination unit 104 refers to the dictionary data and context data stored in the server 20 when determining the control request. In addition, the operation object device determination unit 104 pre-sets each operation object device as the first detection mode object device or the second detection mode object device according to the operation object device. The operation object device determination unit 104 determines which of the first and second detection mode object devices the determined operation object device is by referring to the object device information, and the object device information predetermines which of the first detection mode object device and the second detection mode object device each operation object device is. The first detection mode object device is a device that becomes the object of the first detection mode based on natural language recognition described later when determining the control request. In addition, the second detection mode object device is a device that becomes the object of the second detection mode based on rule base recognition described later when determining the control request. Moreover, the information indicating "which of the first and second detection mode object devices the operation object device is" determined by the operation object device determination unit 104 (hereinafter also referred to as determination information) is output to the display 12 and / or the audio device 13 via the communication unit 107 and the device control unit 11 on the vehicle 1 side. That is, the determination information indicating whether the determined operation object device is the first detection mode object device or the second detection mode object device is displayed on the display 12 in the form of an image including an image and text, or is output by the audio device 13 in the form of voice. In addition, the determination information of the operation target device set by the operation target device identification unit 104 is output to the detection mode setting unit 105 .

[0037] The detection mode setting unit 105 sets the detection mode for determining the control request corresponding to the voice data converted by the voice data conversion unit 102 according to the determination information of the operation object device determined by the operation object device determination unit 104. Specifically, the detection mode setting unit 105 sets the detection mode to one of the first detection mode based on natural language recognition described later and the second detection mode based on rule base recognition described later according to the determination information of the operation object device. The information of the detection mode set by the detection mode setting unit 105 is output to the display 12 and / or the audio device 13 via the communication unit 107 and the device control unit 11 on the vehicle 1 side. That is, the information indicating which of the first detection mode and the second detection mode the current detection mode is set to (hereinafter also referred to as "detection mode setting information") is displayed on the display 12 in the form of an image including an image or text, or is output by the audio device 13 in the form of voice. In addition, the detection mode setting information indicating the detection mode set by the detection mode setting unit 105 is output to the control request determination unit 106.

[0038] The control request determination unit 106 analyzes the voice data converted by the voice data conversion unit 102 based on the detection mode set by the detection mode setting unit 105 to determine the content of the control request for the operation object device. In addition, the control request determination unit 106 refers to the dictionary data and context data stored in the server 20 when determining the control request, similarly to the operation object device determination unit 104. Moreover, the control request determination unit 106 sends the content of the control request for the determined operation object device to the device control unit 11 via the communication unit 107 on the vehicle 1 side. The device control unit 11 controls the operation object device based on the received information of the operation object device and the content of the control request for the operation object device.

[0039] Next, use Figure 2 The flowchart shown explains the process of the speech recognition method using the speech recognition device 100.

[0040] First, in step S1, the voice acquisition unit 101 on the vehicle 1 side of the voice recognition device 100 acquires the voice uttered by the user. Specifically, the voice acquisition unit 101 detects the user's speech by frequency, and when the user's speech is detected, acquires voice information including the voice uttered by the user as a voice recognition object. In this way, the voice acquisition unit 101 acquires voice information only during the period when the user is speaking, thereby reducing the control burden of the voice recognition device 100 compared to the case where voice information is always acquired. In addition, in the case where a specific keyword indicating the start of instruction by voice is predetermined, the voice acquisition unit 101 can also acquire voice information as a voice recognition object just after the user says the specific keyword.

[0041] Next, in step S2 , the speech data conversion unit 102 on the server 20 side of the speech recognition device 100 converts the acquired speech into speech data that can be recognized as text.

[0042] And then, in step S3, the operation object device determination unit 104 analyzes the voice data converted by the voice data conversion unit 102 to determine the operation object device as the control object. The operation object device determination unit 104 analyzes the voice data through natural language recognition to determine the operation object device. Specifically, when the user says "open the sunroof", the operation object device determination unit 104 refers to the dictionary data and context data to determine the operation object device as the sunroof opening and closing device 6b. In addition, when the user says the instruction by voice in the form of "open the upper window" or "open the roof", the operation object device determination unit 104 also determines that the "upper window" or "roof" in this case refers to the sunroof, and determines the operation object device as the sunroof opening and closing device 6b. That is, the operation object device determination unit 104 can determine the operation object device even if the voice data based on the voice uttered by the user is inconsistent with the pre-defined control request text.

[0043] Next, in step S4, the operation target device identification unit 104 determines whether the operation target device identified in step S3 is a first detection mode target device. Figure 3 A method for determining whether an operation target device is a first detection mode target device or whether the operation target device is not a first detection mode target device but a second detection mode target device will be described.

[0044] exist Figure 3 In the example shown, the onboard devices of vehicle 1 are pre-classified into three categories according to the level of operation importance. Operation importance refers to the degree of influence of the control of the onboard device on the driving state of vehicle 1 or the user riding in vehicle 1. Therefore, the higher the operation importance of the onboard device, the greater the influence on the driving of vehicle 1, and therefore the greater the requirement to make the possibility of misrecognition of control requests as close to 0 as possible. In other words, the higher the operation importance of the operation object device, the more it is required that the control request determined by the voice recognition device 100 is accurately and reliably consistent with the actual instruction made by the user's voice.

[0045] Specifically, the navigation device 14, the communication device 15, the audio device 13, the display 12, the interior light 3, the air conditioner 4, the seat massage device 33a, and the seat heater 33b are classified as the operation importance "1 (Low)". It is considered that these accessory devices have a low impact on driving, so these accessory devices are classified as the lowest operation importance. On the other hand, the seat position adjustment device 33c is classified as the operation importance "2 (Middle)". That is, the position and height of the seat where the user sits will affect the user's driving of the vehicle 1, so the operation importance of the seat position adjustment device 33c is higher than that of the seat massage device 33a and the seat heater 33b. In addition, the travel control device 31, the headlight 32, the door and window locking device 5a, the door and window opening and closing device 5b, the sunroof locking device 6a, the sunroof opening and closing device 6b, the tailgate locking device 7a, the tailgate opening and closing device 7b, the trunk locking device 34a, and the trunk opening and closing device 34b have a particularly large impact on the travel of the vehicle 1 or the user, and are therefore classified as the operation importance "3 (High)". The operation object device determination unit 104 determines that the vehicle-mounted device classified as the operation importance "1 (Low)" and the vehicle-mounted device classified as the operation importance "2 (Medium)" are the first detection mode object devices. In addition, the operation object device determination unit 104 determines that the vehicle-mounted device classified as the operation importance "3 (High)" is the second detection mode object device.

[0046] Furthermore, although the distinction between the first detection mode target device and the second detection mode target device is set in advance for each vehicle-mounted device, the operation target device determination unit 104 can appropriately change the setting of the first detection mode target device and the second detection mode target device.

[0047] Then, if Figure 2As shown, if it is determined in step S4 that the operation object device is the first detection mode object device, then in step S5, the detection mode setting unit 105 sets the detection mode to the first detection mode. Then, in step S6, the control request determination unit 106 determines whether the user's instruction content can be read from the voice data converted in step S2 by natural language recognition. Whether the control request determination unit 106 can analyze the voice data converted by the voice data conversion unit 102 based on natural language recognition with reference to the dictionary data and the context data to determine the content of the control request, it is determined whether the instruction content can be read. In the case where the instruction cannot be read from the voice data, in step S12, the voice recognition device 100 listens to the user again. Regarding the re-listening, the user is prompted by displaying it on the display 12 in the form of an image containing characters, or by voice guidance by the audio device 13. On the other hand, in the case where the instruction content can be read from the voice data, in step S7, the control request determination unit 106 determines the control request for the operation object device. That is, the control request determination unit 106 refers to the dictionary data and the context data and interprets the voice data converted by the voice data conversion unit 102 based on natural language recognition to determine the content of the control request. In step S11, the determined control request is sent to the device control unit 11 of the vehicle 1 to control the operation object device, and the process ends.

[0048] If it is determined in step S4 that the operation object device is not the first detection mode object device, that is, it is determined that the operation object device is the second detection mode object device, then in step S8, the detection mode setting unit 105 sets the detection mode to the second detection mode. Moreover, in step S9, it is determined whether the voice data converted by step S2 is consistent with the predefined control request text. If it is inconsistent, in step S12, the user is listened to again. The re-listening to the user is a process of proposing or requesting to re-enter the voice (speak the instructions). At this time, the control request determination unit 106 outputs the guidance information to the display 12 or the audio device 13 via the communication unit 107 and the device control unit 11 of the vehicle 1, and the guidance information guides the user to re-indicate (voice input / speak) in a manner that the voice data is consistent with the control request text.

[0049] On the other hand, when the voice data is consistent with the control request text, in step S10, the control request determination unit 106 determines the control request corresponding to the control request text by using the second detection mode identified based on the rule base. In step S11, the determined control request is sent to the device control unit 11 via the communication unit 107 of the vehicle 1 to control the operation object device, and the process ends.

[0050] In addition, the judgment of whether the voice data in the rule base recognition is consistent with the control request text is performed by comparing the entire voice data (text data) based on the voice input in step S1 with the control request text. The entire voice data is the voice data based on the voice input from the beginning of the speech to the end of the speech. In other words, the control request determination unit judges whether the voice data based on a series of voices from the beginning of the speech to the end of the speech is consistent with the control request text.

[0051] Here, use Figure 4 The difference between the first detection mode based on natural language recognition and the second detection mode based on rule base recognition is explained. Figure 4 , the difference between the first detection mode and the second detection mode is shown when the control target device is an air conditioner and the content of the control request is "I want to turn on the power of the air conditioner". Figure 3 In the example shown, the air conditioner 4 is specified as the first detection mode target device, but in Figure 4 In the example shown, for convenience of explanation, it is assumed that both the first detection mode and the second detection mode can be applied to determining the control request for the air conditioner 4 .

[0052] like Figure 4 As shown, when the control target device is an air conditioner and the content of the control request is "want to turn on the power of the air conditioner", the control request text corresponding to the content of the control request is pre-defined as "turn on the air conditioner".

[0053] In the second detection mode based on rule base recognition, when it is determined that the voice data based on the voice uttered by the user is consistent with the control request text, a control request for the air conditioner as the operation target device is determined (refer to Figure 2 Step S10). That is, Figure 4 In the example shown, when the detection mode setting unit 105 sets the detection mode to the second detection mode, only when the user says "turn on the air conditioner" according to the defined control request text, the device control unit 11 turns on the power of the air conditioner 4. Therefore, when the user says something inconsistent with the control request text, for example, when the user says "turn on the air conditioner", "Hmm...I want to turn on the air conditioner", or "It will be cool if the air conditioner is turned on", the device control unit 11 does not turn on the power of the air conditioner 4, and maintains the off state.

[0054] In the first detection mode based on natural language recognition, in addition to the case where the user says "turn on the air conditioner" as the control request text, the device control unit 11 also turns on the power of the air conditioner 4 when, for example, the user says "turn on the air conditioner" or "hmm... I want to turn on the air conditioner". Specifically, when the user says "turn on the air conditioner", the control request determination unit 106 separates the voice data into "air conditioner" and "turn on", and refers to the dictionary data to determine that "turn on" is a synonym of "turn on", thereby determining the control request of "I want to turn on the power of the air conditioner". In addition, the dictionary data used in the first detection mode stores a registered word and other words with the same meaning as the word as synonyms. In addition, when the user says "hmm... I want to turn on the air conditioner", the control request determination unit 106 determines that "hmm" is an interjection and a meaningless word, and determines that "hmm... I want to turn on the air conditioner" is voice data indicating the same request as "turn on the air conditioner". Thus, the control request determination unit 106 determines the control request “I want to turn on the power of the air conditioner” based on the voice data “Hmm…I want to turn on the air conditioner”.

[0055] That is, regarding the range of voice data that can be recognized corresponding to each control request, the range in the second detection mode based on rule base recognition is smaller than the range in the first detection mode based on natural language recognition. That is, the second detection mode based on rule base recognition only determines control requests defined with control request texts, so the range of voice data that can determine the content of the control request is limited according to the number of defined control requests. In addition, the second detection mode based on rule base recognition only determines control requests consistent with the defined control request texts, so the voice data that can be determined by the second detection mode (that can be input by the second detection mode) is limited. That is, the degree of freedom of voice input when the detection mode is set to the second detection mode is lower than the degree of freedom of voice input when the detection mode is set to the first detection mode. Therefore, the second detection mode has stricter conditions for determining control requests than the first detection mode.

[0056] In other words, regarding the range of voice data that can be recognized corresponding to each operation object device and control request, the range in the first detection mode based on natural language recognition is larger than the range in the second detection mode based on rule base recognition. That is, the first detection mode has looser conditions for determining the operation object device and the control request than the second detection mode.

[0057] On the other hand, when the user says "It will be cool if I turn on the air conditioner", the control request determination unit 106 analyzes the combination and order of words based on the dictionary data and the context data, and infers that the instruction given by the user's voice means "If the air conditioner is turned on, the room temperature of the vehicle will be lower than the appropriate temperature, so I don't want to turn on the air conditioner". Therefore, the control request determination unit 106 does not send an instruction to turn on the power of the air conditioner 4 to the device control unit 11 of the vehicle 1. That is, when the user says "It will be cool if I turn on the air conditioner", regardless of whether the detection mode is set to the first detection mode or the second detection mode, the power of the air conditioner 4 is maintained in the off state and is not turned on.

[0058] Furthermore, the control request determination unit 106 of the speech recognition device 100 can learn the pattern of the correspondence relationship between the speech data and the actual instruction content of the user, and can periodically update the dictionary data and the context data.

[0059] According to the above, the speech recognition device 100 and the speech recognition method using the speech recognition device 100 according to the present embodiment set a detection mode for determining a control request corresponding to speech data according to the operation object device, and determine the control request based on the set detection mode. Therefore, the speech recognition device 100 can determine the control request based on the instruction made by the user's voice in the appropriate detection mode corresponding to each operation object device.

[0060] In addition, when the operation object device determined to be the first detection mode object device is the first detection mode object device, the voice recognition device 100 sets the detection mode to the first detection mode for analyzing the voice data by natural language recognition, and when the operation object device determined to be the second detection mode object device is the second detection mode for analyzing the voice data by voice recognition of a method different from natural language recognition. That is, the voice recognition device 100 can set the detection mode to one of the first detection mode using natural language recognition and the second detection mode using voice recognition of a method different from natural language recognition according to the operation object device. In addition, regarding the range of voice data that can be recognized corresponding to each control request, the range in the second detection mode is smaller than the range in the first detection mode, and the range in the second detection mode is limited. Therefore, the voice recognition device 100 can distinguish between the first detection mode using natural language recognition based on the acceptance of fuzzy instructions (instructions with high degrees of freedom) through voice, and the second detection mode based on voice recognition with a lower degree of freedom than natural language recognition but high accuracy according to the operation object device. In addition, when natural language recognition is used to analyze voice data, sometimes the number of object words that the control request determination unit 106 should recognize increases, or the voice data contains words with multiple meanings. Therefore, compared with the case where the detection mode is set to the second detection mode, the possibility of misrecognition of the control request becomes higher when the detection mode is set to the first detection mode. Thus, for example, in the case where a control instruction is issued to an operation object device with high operation importance and not suitable for ambiguous instructions by voice, the voice recognition device 100 can select the second detection mode that can expect high-accuracy voice recognition as the detection mode. On the other hand, in the case where a control instruction is issued to an operation object device with not too high operation importance, the voice recognition device 100 can select the first detection mode that has a large range of voice recognition and a high degree of freedom of the voice allowed to be input.

[0061] In addition, when the detection mode is set to the second detection mode, the voice recognition device 100 determines whether the voice data is consistent with a predefined control request text. Moreover, when the voice recognition device 100 determines that the voice data is consistent with the control request text, it determines the control request corresponding to the control request text. That is, the second detection mode is a detection mode that analyzes voice data through rule base recognition. Therefore, for example, when determining a control request for an operation object device with high operation importance and not suitable for voice recognition based on natural language recognition, the voice recognition device 100 can select the second detection mode that analyzes voice data through rule base recognition. Therefore, when the voice recognition device 100 determines a control request for an operation object device that is not suitable for voice recognition based on natural language recognition, the accuracy of voice recognition can be improved, thereby preventing misrecognition of instructions made through the user's voice.

[0062] In addition, when the detection mode is set to the second detection mode and the voice data based on the instruction by voice is not consistent with the control request text, the voice recognition device 100 guides the user to give the instruction by voice again. Thus, when the detection mode is set to the second detection mode, the voice recognition device 100 can prompt the user to give the instruction by voice in a way that can perform rule base recognition.

[0063] Furthermore, when the operation object device determination unit 104 determines that the operation object device is the travel control device 31 for controlling the travel of the vehicle 1, the operation object device is determined as the second detection mode object device. The control of the driving actuator, the steering actuator, and the brake actuator of the vehicle 1 by the travel control device 31 has a great influence on the travel state of the vehicle 1, and therefore the operation importance of the travel control device 31 is particularly high. Therefore, by determining that the travel control device 31 is the second detection mode object device by the operation object device determination unit 104, the control request determination unit 106 can reliably and accurately determine the control request for the travel control device 31. Therefore, the stability of the travel state of the vehicle 1 can be maintained.

[0064] In addition, when the operation object device determination unit 104 determines that the operation object device is a locking device of the trunk, door or door window of the vehicle 1, or an opening and closing device of the trunk, door or door window, the operation object device determination unit 104 determines that the operation object device is a second detection mode object device. That is, the operation object device determination unit 104 sets the door and window locking device 5a, the door and window opening and closing device 5b, the sunroof locking device 6a, the sunroof opening and closing device 6b, the back door locking device 7a, the back door opening and closing device 7b, the trunk locking device 34a or the trunk opening and closing device 34b as the second detection mode object device in advance. Whether the trunk, door or door window of the vehicle 1 is locked or whether the trunk, door or door window is opened has a great influence on the driving state of the vehicle 1 and the user riding in the vehicle 1. Therefore, the operation importance of the door and window locking device 5a, the door and window opening and closing device 5b, the sunroof locking device 6a, the sunroof opening and closing device 6b, the back door locking device 7a, the back door opening and closing device 7b, the trunk locking device 34a, and the trunk opening and closing device 34b is particularly high. By determining these locking devices and opening and closing devices as the second detection mode object devices by the operation object device determination unit 104, the control request determination unit 106 can reliably and accurately determine the control request for the door and window control device 5, the sunroof control device 6, the back door control device 7, or the trunk control device 34. Therefore, the stability of the driving state of the vehicle 1 can be maintained.

[0065] When the operation object device determination unit 104 determines that the operation object device is an environment adjustment device for adjusting at least one of the environment conditions including the temperature, light quantity and volume inside the vehicle 1, that is, the interior light 3, air conditioner 4 or audio device 13, the operation object device is determined as the first detection mode object device. The interior light 3, air conditioner 4 or audio device 13 has less influence on the driving state of the vehicle 1 and the user than the driving control device 31 or the locking device or opening and closing device of the door, etc. Therefore, if Figure 3 As shown, the operation importance of the interior light 3, air conditioner 4, and audio device 13 is "1 (low)". Therefore, the interior light 3, air conditioner 4, or audio device 13 is determined as the first detection mode object device by the operation object device determination unit 104, and the control request determination unit 106 can use the first detection mode with a high degree of freedom of the allowed input voice to determine the control request for these devices. As a result, the user can use various speech patterns to more sensibly issue instructions for controlling the on / off of the interior light 3, air conditioner 4, or audio device 13, or the light volume of the interior light 3, the air volume of the air conditioner 4, or the set volume of the audio device 13.

[0066] When the operation object device determination unit 104 determines that the operation object device is a seat control device 33 for controlling the state of a seat provided inside the vehicle 1, that is, a seat massage device 33a, a seat heater 33b, or a seat position adjustment device 33c, the operation object device is determined as a first detection mode object device. The seat control device 33 has less influence on the driving state of the vehicle 1 and the user than the driving control device 31 or a locking device or an opening and closing device such as a door. Therefore, if Figure 3 As shown in FIG. 1 , the operation importance of the seat massage device 33a and the seat heater 33b is "1 (low)", and the operation importance of the seat position adjustment device 33c is "2 (medium)". Therefore, the seat massage device 33a, the seat heater 33b, or the seat position adjustment device 33c is determined as the first detection mode object device by the operation object device determination unit 104, and the control request determination unit 106 can determine the control request for each seat control device 33 using the first detection mode with a high degree of freedom of the voice input allowed. As a result, the user can more intuitively issue instructions for controlling the switching of the massage intensity of the seat massage device 33a, the temperature of the seat surface heated by the seat heater 33b, or the height and position of the seat adjusted by the seat position adjustment device 33c, etc., through various speech modes (speech modes with a high degree of freedom).

[0067] When the operation target device determining unit 104 determines that the operation target device is an information providing device that provides information to the user, that is, the display 12 or the audio device 13, the operation target device is determined as the first detection mode target device. The display 12 or the audio device 13 has less influence on the driving state of the vehicle 1 and the user than the driving control device 31 or the locking device or opening and closing device of the door, etc. Figure 3 As shown, the operation importance of the display 12 or the audio device 13 is "1 (low)". Therefore, the display 12 or the audio device 13 is determined as the first detection mode object device by the operation object device determination unit 104, and the control request determination unit 106 can determine the control request using the first detection mode in which the degree of freedom of the voice input is high. As a result, the user can more intuitively issue instructions for controlling the on / off of the display 12 or the audio device 13, the brightness of the screen of the display 12, the size of the characters, or the set volume of the audio device 13, etc., through various speech patterns.

[0068] The audio device 13 is an environment adjustment device that adjusts the environment inside the vehicle 1 and is also an information providing device. The information providing device is not limited to the display 12 and the audio device 13, and may include an instrument panel (not shown) and the like.

[0069] In addition, Figure 2In step S3 shown, the operation object device determination unit 104 analyzes the voice data based on natural language recognition to determine the operation object device. Thus, even if the user issues an instruction by voice in a language inconsistent with the control request text, the operation object device determination unit 104 can still determine the operation object device.

[0070] Furthermore, in the present embodiment, the voice data conversion unit 102 , the operation target device determination unit 104 , the detection mode setting unit 105 , and the control request determination unit 106 are provided in the server 20 , but some of these functions may be provided in the vehicle 1 .

[0071] Second Implementation Method

[0072] exist Figure 5 FIG. 2 shows a structure of a vehicle 1 including a speech recognition device 200 according to the second embodiment. Figures 1 to 4 The same reference numerals as those in the drawings denote the same or similar components or steps, and thus detailed descriptions thereof will be omitted.

[0073] like Figure 5 As shown, the vehicle 1 has a navigation system 110, which controls one or more of the display 12, the audio device 13, the navigation device 14, and the communication device 15, or all of them by voice input. The navigation system 110 has a voice recognition device 200, which accepts control requests for various devices mounted on the vehicle 1 based on instructions given by voice from a user riding in the vehicle 1. The voice acquisition unit 101, the voice data conversion unit 102, the operation object device determination unit 104, the detection mode setting unit 105, the control request determination unit 106, and the communication unit 107 that constitute the voice recognition device 200 are all provided in the vehicle 1. The detection mode setting unit 105 and the control request determination unit 106 can exchange data or signals with the device control unit 11 via the communication unit 107.

[0074] In addition, the speech recognition device 200 Figure 2 The voice recognition method shown in the figure determines the control request for the operation target device in the same process.

[0075] Based on the above, similarly to the speech recognition device 100 involved in the first embodiment, the speech recognition device 200 involved in this embodiment sets a detection mode for determining a control request corresponding to speech data according to the operation object device, and determines the control request based on the set detection mode. Therefore, the speech recognition device 200 can determine the control request based on the instruction made by the user's voice in an appropriate detection mode corresponding to each operation object device, similarly to the speech recognition device 100. In addition, since the speech recognition device 200 is mounted on the vehicle 1, it is possible to determine the control request for the operation object device based on the instruction made by the user's voice regardless of the communication environment.

[0076] 《Third Implementation Mode》

[0077] Reference Figures 6 to 10 A speech recognition device 300 according to a third embodiment and a speech recognition method using the speech recognition device 300 will be described.

[0078] like Figure 6 As shown, the server 20 has a voice collection environment determination unit 108. The voice collection environment determination unit 108 determines whether the voice collection environment for the voice acquisition unit 101 to acquire voice is good. In addition, the vehicle 1 has a vehicle speed sensor 2 for detecting the driving speed of the vehicle 1. The vehicle speed sensor 2 detects the driving speed of the vehicle 1, for example, based on the rotation speed (wheel speed) of the wheels of the vehicle 1. In addition, the vehicle 1 has a wind speed sensor 8 for detecting the wind speed around the vehicle 1. In addition, a vehicle information acquisition unit 103 for acquiring status information of the vehicle 1 is provided in the vehicle 1. The vehicle information acquisition unit 103 is connected to the vehicle speed sensor 2 or the wind speed sensor 8 wirelessly or by wire. In addition, the vehicle information acquisition unit 103 is also connected to the voice acquisition unit 101 and the communication unit 107 wirelessly or by wire.

[0079] Furthermore, the vehicle information acquisition unit 103 constitutes a moving body information acquisition unit.

[0080] The display 12, the audio device 13, the navigation device 14, the call device 15, the device control unit 11, the voice acquisition unit 101, the vehicle information acquisition unit 103 and the communication unit 107 provided in the vehicle 1, and the voice collection environment determination unit 108, the voice data conversion unit 102, the operation object device determination unit 104, the detection mode setting unit 105 and the control request determination unit 106 of the server 20 constitute the navigation system 210. In addition, the voice acquisition unit 101, the vehicle information acquisition unit 103 and the communication unit 107 provided in the vehicle 1, and the voice collection environment determination unit 108, the voice data conversion unit 102, the operation object device determination unit 104, the detection mode setting unit 105 and the control request determination unit 106 of the server 20 constitute the voice recognition device 300. That is, the voice recognition device 300 constitutes a part of the navigation system 210. In addition, the voice recognition device 300 receives an instruction given by a voice from a user, and controls one or more or all of the in-vehicle devices through the device control unit 11. That is, the navigation system 210 includes the voice recognition device 300 that receives control requests for various devices mounted on the vehicle 1 based on instructions given by voice from a user riding in the vehicle 1. The voice recognition device 300 is a voice recognition system.

[0081] Here, the vehicle information acquisition unit 103 acquires the state information of the vehicle 1 via the vehicle speed sensor 2, the wind speed sensor 8, the voice acquisition unit 101, and the communication unit 107. The state information of the vehicle 1 is information about factors that affect the voice environment in the interior of the vehicle 1. For example, the state information of the vehicle 1 is the size of the noise in the interior of the vehicle 1, the opening and closing information of the openings such as doors, windows, sunroofs, or tailgates, the driving speed of the vehicle 1, the wind speed around the vehicle 1, the set volume of the audio device 13, the air volume of the air blown out from the air conditioner 4, etc. The size of the noise in the interior of the vehicle 1 is the size (for example, a measured value) of the noise detected by the voice acquisition unit 101, that is, the size of the sound other than the voice spoken by the user, and the size of the noise is calculated in dB (decibel). In addition, the vehicle information acquisition unit 103 acquires the set volume of the audio device 13, the air volume of the air blown out from the air conditioner 4, the opening and closing information of doors, windows, sunroofs, or tailgates, etc. as the state information of the vehicle 1 via the communication unit 107 and the device control unit 11. In addition, the vehicle information acquisition unit 103 can also calculate the noise level inside the vehicle 1, for example, the interior of the vehicle 1, based on the magnitude of the noise inside the vehicle 1 and the set volume of the audio device 13, and acquire the calculated noise level as the state information of the vehicle 1. Specifically, the noise level is calculated by adding the score assigned according to the magnitude of the noise inside the vehicle 1 and the score assigned according to the set volume of the audio device 13. The noise level can be calculated using the measurement value obtained by the microphone installed in the vehicle cabin of the vehicle 1, or using the measurement value obtained by the microphone installed in the vehicle body of the vehicle 1 or the vibration sensor.

[0082] The voice collection environment determination unit 108 provided in the server 20 determines whether the voice collection environment in the interior of the vehicle 1 is good or not based on the state information of the vehicle 1 acquired by the vehicle information acquisition unit 103. The voice collection environment in the interior of the vehicle 1 is an environment used by the voice acquisition unit 101 of the voice recognition device 300 to acquire the voice uttered by the user. The better the voice collection environment in the interior of the vehicle 1, the less noise (sounds other than the user's speaking voice) contained in the voice acquired by the voice acquisition unit 101, and the higher the possibility that the voice recognition device 300 can accurately recognize the instructions made by the user's voice. On the other hand, the worse the voice collection environment in the interior of the vehicle 1, the more noise contained in the voice acquired by the voice acquisition unit 101, and the higher the possibility that the voice recognition device 300 mistakenly determines the instructions made by the user's voice, that is, misrecognition. In addition, the recognition process of the instructions made by the user's voice includes the determination process of the operation object device to be controlled and / or the determination process of the content of the control request for the operation object device.

[0083] Next, use Figure 7The process of the speech recognition method using the speech recognition device 300 is described with reference to the flowchart shown.

[0084] like Figure 7 As shown, if it is determined in step S4 that the operation target device is the first detection mode target device, then in step S13, the vehicle information acquisition unit 103 acquires the state information of the moving body, that is, the state information of the vehicle. Then, in step S14, the voice collection environment determination unit 108 determines whether the voice collection environment is good based on the state information of the vehicle.

[0085] In addition, the processing of step S13 can be performed in parallel with the processing of steps S1 to S4. The processing of step S13 can be performed continuously or periodically after the start of this control process. In addition, the vehicle information acquisition unit 103 can continuously or periodically acquire the status information of the vehicle from before the start of this control process, and in the processing of step S13, the status information of the vehicle after voice acquisition is extracted and acquired from the status information acquired continuously or periodically in advance.

[0086] Here, use Figure 8 The following describes the criteria for determining whether the voice collection environment is good based on the vehicle status information.

[0087] Whether the voice collection environment is good or not is determined based on the misrecognition level corresponding to the status information of the vehicle 1. The misrecognition level is an indicator indicating the possibility that the voice recognition device 300 cannot accurately recognize the voice uttered by the user and thus cannot determine the correct control request. Figure 8 In the example shown, the misrecognition level is expressed as a quantitative index value. That is, the higher the index value of the misrecognition level, the more noise there is in the interior of the vehicle 1, so that the voice data conversion unit 102 cannot accurately convert the voice into voice data, and the possibility of misrecognition of the instruction made by the user's voice becomes higher. Therefore, the higher the index value of the misrecognition level, the worse the voice collection environment. Figure 8 In the table shown, the added score of the misrecognition level (index value) corresponding to the state information of the vehicle 1 is specified. Figure 8 As an example of a calculation method of the misrecognition level, it is possible to predefine the added points and the added items according to the characteristics of the vehicle, the characteristics of the voice recognition device, and the like.

[0088] like Figure 8As shown, the added score of the misrecognition level when the sunroof W1 or the door window W2 of the vehicle 1 is in the open state is 1 point. In addition, the added score of the misrecognition level when the air volume of the air conditioner 4 is above the predetermined reference air volume is 2 points. In addition, the added score of the misrecognition level when the sunroof W1 or the door window W2 is in the open state and the driving speed of the vehicle 1 is above 20 km / h and below 80 km / h is 2 points. In addition, the added score of the misrecognition level when the sunroof W1 or the door window W2 is in the open state and the driving speed of the vehicle 1 is above 80 km / h is 3 points. And, the added score of the misrecognition level when the noise level in the room of the vehicle 1 is above the predetermined reference noise level is 3 points.

[0089] The sunroof W1 and the door window W2 of the vehicle 1 constitute an opening portion of the vehicle 1. The opening portion of the vehicle 1 is not limited to the sunroof W1 and the door window W2, and may be, for example, a door of the vehicle 1.

[0090] By using the status information of the vehicle 1 Figure 8 The misrecognition level (index value) of the voice collection environment is calculated by adding the scores shown on the right side of the table. For example, for the misrecognition level when the sunroof W1 is open and the vehicle speed is 20 km / h or more and less than 80 km / h, 1 point and 2 points are added to calculate 3 points. In addition, for the misrecognition level when the sunroof W1 is open, the air conditioning air volume is greater than the reference air volume, and the vehicle speed is 80 km / h or more, 1 point, 2 points, and 3 points are added to calculate 6 points.

[0091] In step S14, the voice collection environment determination unit 108 determines the reference level of the misrecognition level as 4 points, and determines that the voice collection environment is good when the misrecognition level is less than 4 points as the reference level, and determines that the voice collection environment is bad when the misrecognition level is 4 points or more. In addition, the score of the reference level can be defined in advance based on vehicle characteristics, characteristics of the voice recognition device, etc. In addition, the score of the reference level can also be defined in advance based on experiments or simulations.

[0092] If it is determined in step S14 that the voice collection environment is good, then in step S5, the detection mode setting unit 105 sets the detection mode to the first detection mode. That is, when the voice collection environment determination unit 108 determines that the voice collection environment is good and the operation object device determination unit 104 determines that the operation object device is a first detection mode object device, the detection mode setting unit 105 sets the detection mode to the first detection mode.

[0093] On the other hand, if it is determined in step S14 that the voice collection environment is poor, then in step S15, the operation object device determination unit 104 changes the setting of the operation object device from the first detection mode object device to the second detection mode object device. That is, when the voice collection environment determination unit 108 determines that the voice collection environment is poor, the operation object device determination unit 104 changes the operation object device that was previously set to the first detection mode object device and resets it to the second detection mode object device. The server 20 outputs information indicating the change in the setting of the operation object device, that is, information including that the setting of the operation object device has been changed from the first detection mode object device to the second detection mode object device, to the display 12 or the audio device 13 via the communication unit 107 and the device control unit 11 of the vehicle 1. Moreover, in step S8, the detection mode setting unit 105 sets the detection mode to the second detection mode.

[0094] Here, in Fig. 9 2 shows an example of a state in which, in step S15 , the operation object device identification unit 104 changes and resets the setting of the operation object device from the first detection mode object device to the second detection mode object device.

[0095] If it is determined in step S14 that the voice collection environment is bad, then Fig. 9 As shown, the navigation device 14, the communication device 15, the audio device 13, the display 12, the interior light 3, the air conditioner 4, the seat massage device 33a, and the seat heater 33b classified as the operation importance "1 (low)" are determined as the second detection mode object devices, and the control request is determined by the second detection mode recognized based on the rule base. In addition, the seat position adjustment device 33c classified as the operation importance "2 (medium)" is also determined as the second detection mode object device, and the control request is determined by the second detection mode recognized based on the rule base.

[0096] In addition, the travel control device 31, headlight 32, door and window locking device 5a, door and window opening and closing device 5b, sunroof locking device 6a, sunroof opening and closing device 6b, tailgate locking device 7a, tailgate opening and closing device 7b, trunk locking device 34a, and trunk opening and closing device 34b, which are classified as the operation importance "3 (high)", do not change the setting. Therefore, if it is determined in step S14 that the voice collection environment is poor, the operation object device is set to the second detection mode object device, and the second detection mode based on rule base recognition is applied in determining the control request for the operation object device.

[0097] According to the above, in the speech recognition device 300 and the speech recognition method using the speech recognition device 300 according to the present embodiment, when the speech collection environment determination unit 108 determines that the speech collection environment is good and the operation object device determination unit 104 determines that the operation object device is the first detection mode object device, the detection mode setting unit 105 sets the detection mode to the first detection mode. Therefore, the speech recognition device 300 applies the first detection mode based on natural language recognition to the first detection mode object device with low operation importance only when the speech collection environment is good, that is, when there is little noise in the speech collection environment, and determines the control request. That is, the speech recognition device 300 is configured to avoid applying the first detection mode based on natural language recognition in a speech collection environment with a lot of noise. Therefore, even in the case of determining the control request by natural language recognition with a high degree of freedom of speech input, the speech recognition device 300 can reduce the possibility of misrecognition according to the speech collection environment, so that the control request for the operation object device can be determined more reliably.

[0098] In addition, when the voice collection environment determination unit 108 determines that the voice collection environment is poor, the operation object device determination unit 104 changes the operation object device that was previously set as the first detection mode object device and resets it to the second detection mode object device. Here, in the case of a poor voice collection environment, that is, a voice collection environment with a lot of noise, the possibility of misrecognition in the determination of the control request becomes high. Therefore, by resetting the operation object device that was previously set as the first detection mode object device to the second detection mode object device and applying the second detection mode identified based on the rule base to determine the control request, it is possible to prevent misrecognition.

[0099] The voice collection environment determination unit 108 obtains the air volume of air blown out from the air conditioner 4 as the air conditioning device as state information. The greater the air volume of the air conditioner 4, the easier it is for noise to be mixed into the voice acquired by the voice acquisition unit 101. Therefore, by obtaining the air volume of air blown out from the air conditioner 4 as state information, the voice collection environment determination unit 108 can more appropriately determine whether the voice collection environment is good.

[0100] In addition, the voice collection environment determination unit 108 obtains the opening and closing state of the sunroof W1 or the door window W2, which is an opening provided in the vehicle 1, and the driving speed of the vehicle 1 when the sunroof W1 or the door window W2 is in the open state as the state information. This is because, when the vehicle 1 is driving with the sunroof W1 or the door window W2 open, the faster the driving speed of the vehicle 1, the greater the noise that enters the interior of the vehicle 1 from the outside, such as wind sound and driving sound of oncoming vehicles. Therefore, by obtaining the opening and closing state of the sunroof W1 or the door window W2 and the driving speed of the vehicle 1 as the state information, the voice collection environment determination unit 108 can more appropriately determine whether the voice collection environment is good.

[0101] Furthermore, the voice collection environment determination unit 108 obtains the opening and closing state of the opening portion provided in the vehicle 1, namely the sunroof W1 or the door window W2, and the wind speed around the vehicle 1 when the sunroof W1 or the door window W2 is in the open state as state information. This is because, when the sunroof W1 or the door window W2 is in the open state, the faster the wind speed around the vehicle 1, the more wind volume blown into the interior of the vehicle 1, and thus the greater the noise. In addition, even when the vehicle 1 is not traveling but stopped, if the sunroof W1 or the door window W2 is in the open state, the wind speed around the vehicle 1 will also affect the voice collection environment in the interior of the vehicle 1. Therefore, by obtaining the opening and closing state of the sunroof W1 or the door window W2 and the wind speed around the vehicle 1 as state information, the voice collection environment determination unit 108 can more appropriately determine whether the voice collection environment is good.

[0102] In addition, the opening portion of the vehicle 1 is not limited to the sunroof W1 or the door window W2 , and may be another window or a door.

[0103] In addition, the voice collection environment determination unit 108 obtains the noise level in the interior of the vehicle 1 as state information. The higher the noise level in the interior of the vehicle 1, the easier it is for noise to be mixed into the voice acquired by the voice acquisition unit 101. Therefore, by obtaining the noise level in the interior of the vehicle 1 as state information, the voice collection environment determination unit 108 can more appropriately determine whether the voice collection environment is good.

[0104] Furthermore, when the operation object device determination unit 104 changes and resets the operation object device previously set as the first detection mode object device to the second detection mode object device, the voice recognition device 300 outputs information indicating the change in the setting of the operation object device to the display 12 or the audio device 13 provided in the vehicle 1. Thus, the user can reliably confirm that the setting of the operation object device has been changed from the first detection mode object device to the second detection mode object device, and the user can easily issue a voice instruction that matches the current setting of the operation object device.

[0105] In addition, the speech recognition method using the speech recognition device 300 according to the present embodiment is not limited to Figure 7 The process shown in Fig.10 Another example of a speech recognition method using the speech recognition device 300 is shown in FIG.

[0106] like Fig.10 As shown, if the operation object device is determined in step S3, then in step S23, the vehicle information acquisition unit 103 acquires the status information of the vehicle. In addition, the processing of step S23 can be performed in parallel with the processing of steps S1 to S4. Moreover, in step S24, the voice collection environment determination unit 108 determines whether the voice collection environment is good based on the status information of the vehicle. If it is determined in step S24 that the voice collection environment is poor, the voice recognition device 300 does not determine whether the operation object device is the first detection mode object device, but sets the detection mode to the second detection mode. That is, when the voice collection environment determination unit 108 determines that the voice collection environment is good, and the operation object device determination unit 104 determines that the operation object device is the second detection mode object device, or when the voice collection environment determination unit 108 determines that the voice collection environment is poor, the detection mode setting unit 105 sets the detection mode to the second detection mode. In this case, when it is determined that the voice collection environment is poor, as Fig. 9 As shown, the operation object devices are all set as the second detection mode object devices, and the second detection mode based on rule base recognition is applied in the determination of the control request for the operation object device. Thus, even in the case where the control request is determined by natural language recognition with a high degree of freedom of voice input, the voice recognition device 300 can reduce the possibility of misrecognition according to the voice collection environment.

[0107] In addition, the vehicle information acquisition unit 103 can acquire the driving state of the vehicle 1 as the state information of the vehicle 1. In this case, when the operation object device determination unit 104 determines that the vehicle 1 is stopped based on the driving state of the vehicle 1, it can change the operation object device that was previously set as the second detection mode object device and reset it to the first detection mode object device. Specifically, Figure 3The headlight 32, the door and window locking device 5a, the door and window opening and closing device 5b, the sunroof locking device 6a, the sunroof opening and closing device 6b, the tailgate locking device 7a, the tailgate opening and closing device 7b, the trunk locking device 34a, and the trunk opening and closing device 34b shown as the second detection mode object device have less influence on the user during the period when the vehicle 1 is stopped. Therefore, the settings of these vehicle-mounted devices can be changed to the first detection mode object device, and the control request can be determined by the first detection mode based on natural language recognition with a high degree of freedom of voice input. That is, during the period when the vehicle 1 is stopped, the operation object devices can be set to the first detection mode object device, and the first detection mode based on natural language recognition can be applied in the determination of the control request for the operation object device. In addition, when the operation object device determination unit 104 changes the operation object device previously set as the second detection mode object device and resets it to the first detection mode object device, the voice recognition device 300 can output information indicating the change of the setting of the operation object device to the display 12 or the audio device 13 provided in the vehicle 1.

[0108] In addition, the speech recognition devices 100, 200, and 300 can set the first detection mode or the second detection mode according to the determined operation object device regardless of whether natural language recognition or rule base recognition is used as the speech recognition method. For example, the first detection mode and the second detection mode can also be detection modes based on natural language recognition. In this case, the range of speech data that can be recognized corresponding to each control request is set to be smaller in the second detection mode than in the first detection mode. Specifically, the first detection mode can be set to a detection mode based on multi-intent natural language recognition, and the second detection mode can be set to a detection mode based on single-intent natural language recognition. In the detection mode based on multi-intent natural language recognition, the control request determination unit 106 can determine two control object devices (air conditioner 4 and door window W2) and control requests for each control object device ("want to turn off the air conditioner" and "want to open the door window") according to a voice instruction containing more than two control requests, such as "turn off the air conditioner, open the window". On the other hand, in the detection mode based on natural language recognition of a single intent, the control request determination unit 106 can determine a control request only when each voice instruction includes one control request, such as “turn off the air conditioner” or “open the window”.

[0109] In addition, the speech recognition apparatus 100, 200, 300 can make the method of natural language recognition in the following manner: Figure 3As shown, the operation object device classified as the operation importance "1 (low)" is different from the operation object device classified as the operation importance "2 (medium)". Specifically, the detection mode setting unit 105 can set a first detection mode based on natural language recognition of multiple intents for the operation object device classified as the operation importance "1 (low)", and set a first detection mode based on natural language recognition of a single intent for the operation object device classified as the operation importance "2 (medium)". In addition, without being limited to this, the detection mode setting unit 105 can appropriately set the detection mode based on natural language recognition having two or more methods according to each first detection mode object device.

[0110] In addition, the moving object is not limited to a vehicle, and may be a train or an airplane.

[0111] The above-mentioned vehicle 1 is equivalent to the mobile body involved in the present invention. The above-mentioned interior lamp 3 is equivalent to the environmental adjustment device involved in the present invention. The above-mentioned air conditioner 4 is equivalent to the environmental adjustment device and the air conditioning device involved in the present invention. The above-mentioned display 12 is equivalent to the information providing device involved in the present invention. The above-mentioned audio device 13 is equivalent to the environmental adjustment device and the information providing device involved in the present invention. The above-mentioned vehicle information acquisition unit 103 is equivalent to the mobile body information acquisition unit of the present invention. The above-mentioned sunroof W1 and door window W2 are equivalent to the opening part of the present invention.

[0112] Description of Reference Numerals

[0113] 100, 300: voice recognition device (voice recognition system); 200: voice recognition device; 1: vehicle (mobile body); 3: interior lights (environmental adjustment device); 4: air conditioner (environmental adjustment device / air conditioning device); 5a: door and window locking device; 5b: door and window opening and closing device; 6a: sunroof locking device; 6b: sunroof opening and closing device; 7a: rear door locking device; 7b: rear door opening and closing device; 12: display (information providing device); 13: audio device (environmental adjustment device / information providing device); information providing device); 31: driving control device; 33: seat control device; 34a: trunk locking device; 34b: trunk opening and closing device; 101: voice acquisition unit; 102: voice data conversion unit; 103: vehicle information acquisition unit (moving body information acquisition unit); 104: operation object device determination unit; 105: detection mode setting unit; 106: control request determination unit; 108: voice collection environment determination unit; W1: sunroof (opening); W2: door window (opening).

Claims

1. A voice recognition device that receives a control request for a device mounted on a mobile body based on an instruction given by a voice uttered by a user, the voice recognition device comprising: A voice acquisition unit, which acquires the voice; a voice data conversion unit, which converts the voice acquired by the voice acquisition unit into voice data; an operation target device determination unit that analyzes the voice data converted by the voice data conversion unit to determine the operation target device to be operated; a detection mode setting unit that sets one of a first detection mode and a second detection mode as a detection mode for determining the control request corresponding to the voice data according to the operation object device determined by the operation object device determining unit, wherein the first detection mode is used to analyze the voice data through natural language recognition, and the second detection mode is used to determine whether the voice data is consistent with a predefined control request text; and a control request determination unit that analyzes the voice data based on the detection mode set by the detection mode setting unit to determine the control request for the operation target device, Wherein, the speech recognition device further comprises: a moving object information acquiring unit that acquires state information of the moving object; and a voice collection environment determination unit that determines whether a voice collection environment for acquiring the voice is good based on the state information of the mobile body acquired by the mobile body information acquisition unit, When the voice collection environment determination unit determines that the voice collection environment is good and the operation target device determination unit determines that the operation target device is a first detection mode target device, the detection mode setting unit sets the detection mode to the first detection mode.

2. The speech recognition device according to claim 1, characterized in that: When the determined operation target device is a second detection mode target device, the detection mode setting unit sets the detection mode to the second detection mode, Regarding the range of the voice data that can be recognized corresponding to each of the control requests, the range in the second detection mode is smaller than the range in the first detection mode.

3. The speech recognition device according to claim 2, characterized in that: When the detection mode setting unit sets the detection mode to the second detection mode, The control request determination unit determines whether the voice data is consistent with a predefined control request text, The control request determination unit determines the control request corresponding to the control request text when determining that the voice data is consistent with the control request text.

4. The speech recognition device according to claim 3, characterized in that: When the detection mode is set to the second detection mode and the voice data based on the instruction by voice does not match the control request text, The control request determination unit outputs guidance information for guiding the user to re-issue the instruction so that the voice data is consistent with the control request text to an information providing device provided in the mobile body.

5. The speech recognition device according to any one of claims 2 to 4, characterized in that: The operation target device identification unit determines that the operation target device is the second detection mode target device when the operation target device is identified as a travel control device for controlling travel of the mobile body.

6. The speech recognition device according to any one of claims 2 to 4, characterized in that: When the operation object device is determined to be a locking device for the trunk, doors or windows of the mobile body, or an opening and closing device for the trunk, doors or windows, the operation object device determination unit determines the operation object device as the second detection mode object device.

7. The speech recognition device according to any one of claims 2 to 4, characterized in that: The operation target device identification unit determines that the operation target device is the first detection mode target device when the operation target device is identified as an environment adjustment device that adjusts an environment state including at least any one of temperature, light amount, and sound volume inside the moving body.

8. The speech recognition device according to any one of claims 2 to 4, characterized in that: The operation target device identification unit determines that the operation target device is the first detection mode target device when the operation target device is identified as a seat control device that controls the state of a seat provided inside the mobile body.

9. The speech recognition device according to any one of claims 2 to 4, characterized in that: The operation target device identification unit determines that the operation target device is the first detection mode target device when it is determined that the operation target device is an information providing device that provides information to the user.

10. The speech recognition device according to claim 1, characterized in that: When the voice collection environment determination unit determines that the voice collection environment is bad, the operation object device determination unit changes and resets the operation object device previously set as the first detection mode object device to the second detection mode object device.

11. The speech recognition device according to claim 1, characterized in that: The moving body information acquisition unit acquires, as the state information, a volume of air blown out from an air conditioning device installed in the moving body.

12. The speech recognition device according to claim 1, characterized in that: The moving body information acquisition unit acquires, as the state information, an open / closed state of an opening provided in the moving body and a travel speed of the moving body when the opening is in an open state.

13. The speech recognition device according to claim 1, characterized in that: The moving object information acquisition unit acquires, as the state information, an open / closed state of an opening provided in the moving object and a wind speed around the moving object when the opening is in an open state.

14. The speech recognition device according to any one of claims 2 to 4, characterized in that: In a case where the operation object device determination unit changes the operation object device that was previously set as the first detection mode object device and resets it to the second detection mode object device, or in a case where the operation object device that was previously set as the second detection mode object device is changed and reset to the first detection mode object device, information indicating the change in the setting of the operation object device is output to an information providing device provided on the mobile body.

15. The speech recognition device according to any one of claims 1 to 4, characterized in that: The operation target device identification unit analyzes the voice data based on natural language recognition to identify the operation target device.

16. The speech recognition device according to any one of claims 1 to 4, characterized in that: The detection mode is set corresponding to the division of the operation object device determined according to the operation importance, wherein the detection mode setting unit sets the operation object device classified into a first category of operation importance to the first detection mode, and sets the operation object device classified into a second category of operation importance to the second detection mode, wherein the operation importance of the first category is lower than the operation importance of the second category.

17. A voice recognition device that receives a control request for a device mounted on a mobile body based on an instruction given by a voice uttered by a user, the voice recognition device comprising: A voice acquisition unit, which acquires the voice; a voice data conversion unit, which converts the voice acquired by the voice acquisition unit into voice data; an operation target device determination unit that analyzes the voice data converted by the voice data conversion unit to determine the operation target device to be operated; a detection mode setting unit that sets one of a first detection mode and a second detection mode as a detection mode for determining the control request corresponding to the voice data according to the operation object device determined by the operation object device determining unit, wherein the first detection mode is used to analyze the voice data through natural language recognition, and the second detection mode is used to determine whether the voice data is consistent with a predefined control request text; and a control request determination unit that analyzes the voice data based on the detection mode set by the detection mode setting unit to determine the control request for the operation target device, The speech recognition device further comprises a moving body information acquisition unit, wherein the moving body information acquisition unit acquires state information of the moving body. The moving body information acquiring unit acquires the running state of the moving body as the state information, The operation object device identification unit changes and resets the operation object device previously set as the second detection mode object device to the first detection mode object device when it is determined based on the traveling state of the moving body that the moving body is stopped.

18. A voice recognition method, comprising: using a voice recognition device, wherein the voice recognition device receives a control request for a device mounted on a mobile body based on an instruction given by a voice uttered by a user, wherein: The speech recognition device performs the following processing: Acquire the voice; Convert the acquired voice into voice data; Analyzing the voice data to determine an operation target device as an operation target; According to the operation object device, setting one of a first detection mode and a second detection mode as a detection mode for detecting the control request corresponding to the voice data, wherein the first detection mode is used to analyze the voice data through natural language recognition, and the second detection mode is used to determine whether the voice data is consistent with a predefined control request text; and The voice data is analyzed based on the set detection mode to determine the control request for the operation object device, The speech recognition device further performs the following processing: Acquiring status information of the mobile object; and determining whether a voice collection environment for acquiring the voice is good based on the acquired state information of the moving object, When it is determined that the voice collection environment is good and the operation target device is determined to be a first detection mode target device, the detection mode is set to the first detection mode.

19. A voice recognition method, which is a method using a voice recognition device, wherein the voice recognition device receives a control request for a device mounted on a mobile body based on an instruction given by a voice uttered by a user, wherein: The speech recognition device performs the following processing: Acquire the voice; Convert the acquired voice into voice data; Analyzing the voice data to determine an operation target device as an operation target; According to the operation object device, setting one of a first detection mode and a second detection mode as a detection mode for detecting the control request corresponding to the voice data, wherein the first detection mode is used to analyze the voice data through natural language recognition, and the second detection mode is used to determine whether the voice data is consistent with a predefined control request text; and The voice data is analyzed based on the set detection mode to determine the control request for the operation object device, The speech recognition device further performs the following processing: Acquiring the state information of the mobile object, acquiring the driving state of the moving body as the state information, When it is determined based on the traveling state of the moving body that the moving body has stopped, the operation target device previously set as the second detection mode target device is changed and reset to the first detection mode target device.

Images