JP3979080B2 - Voice recognition device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
  The present invention relates to a voice recognition device for easily executing a vehicle control command for changing a running state of a vehicle while maintaining a feeling of operation by using voice recognition and an operation input device. The present invention relates to a speech recognition apparatus suitable for an autonomous traveling vehicle that can autonomously control braking and steering and can receive a command from a driver.
[0002]
【TraditionalTechnology]
  As an introduction to the vehicle control of the voice recognition interface, for example, there is one described in Japanese Patent Application Laid-Open No. 2000-320657 by Toyota Motor. In this voice recognition interface, the running characteristics are set to a predetermined mode by voice recognition. By setting, for example, control of the opening degree of the electronic throttle, switching of the driving method between two-wheel drive and four-wheel drive, and the like are performed. In addition, for example, when it is detected that the road surface μ is clearly high and it is determined that it is not snowing, even if the voice command from the driver is recognized as “snow” by voice recognition, such control is not appropriate, It is also determined that the recognition has been erroneously recognized and the processing is stopped.
[0003]
  In the case of such control of only the vehicle characteristics, even if the vehicle characteristics are changed due to misrecognition, there is no significant influence on the driving such as changing the course direction. The driver's feeling is not greatly impaired. However, when such voice recognition control is introduced in a vehicle that autonomously controls acceleration / deceleration and steering, control in an unintended direction is added, or conversely, the determination is made too strict and intended control is performed. Otherwise, the operability will be significantly impaired. For this reason, if we always use a confirmation method that repeats commands that are generally used by voice, and checks on the monitor screen or touch panel screen, it will take time from the time the voice command is given to the execution. It gets worse. Therefore, the degree of coincidence (here called accuracy) of how much the voice command recognized by the vehicle matches the command intended by the driver is improved, and high accuracy is obtained before the control command is put into execution. It is necessary for the driver to be able to confirm what is being done without losing operability due to time delay.
[0004]
  Accordingly, an object of the present invention is to use a command input intended by the driver by using the operation input and road information in addition to the voice input in order to easily and quickly execute the voice command by the driver without impairing the operational feeling. It is to provide a voice recognition device that enables quick voice recognition by using a direct operation input from a driver while maintaining a high “accuracy” indicating the degree of coincidence with the voice.
[0005]
[Means for Solving the Problems]
  The invention described in claim 1 of the present invention
  Voice input recognition means for recognizing voice input information;
  Operation input recognition means for recognizing operation input information;
  The voiceinputVehicle control command generation means for generating a vehicle control command based on the voice input information recognized by the recognition means and the command recognized by the operation input recognition means;
  Vehicle control means for controlling acceleration / deceleration and steering of the vehicle based on the vehicle control command generated by the vehicle control command generation means;,
  The branch information to which the vehicle can move, the road structure such as the lane width obtained from the map information, the road condition such as the congestion degree obtained from the external recognition sensor and the infrastructure information, and the distance to each branch of the branch information Road information recognition means for recognizing road information including at least one of distance coefficients that are weights according to
  The vehicle control command generating means, comprising: an accuracy calculating means for calculating a probability that the vehicle in the branch information is transferred to each branch to which the vehicle can move based on the voice input information, operation input information, and road information. Generates a command for making a transition to the branch with the highest accuracy as the vehicle control command.The speech recognition apparatus is characterized by the above.
[0006]
  Claims of the invention2According to the invention described in claim 1, in the voice recognition device according to claim 1, the voice input recognition means includes, in the voice input information, a vehicle control action target, an action with respect to the target, and a direction of the action. The speech recognition apparatus is characterized by recognizing at least one.
[0007]
  Claims of the invention3The invention described in claim 11 or 2The speech recognition apparatus according to claim 1, wherein the vehicle control command generation unit determines whether or not to immediately send the vehicle control command to the vehicle control unit according to the accuracy.
[0008]
  Claims of the invention4The invention described in claim 1Any one of 1 to 3In the voice recognition device according to claim 1, when the accuracy calculation means calculates the accuracy of shifting to each branch to which the vehicle can move, it is missing in the voice input information, operation input information, and road information. The speech recognition apparatus is characterized in that the accuracy is calculated based on the remaining information even if there is information.
[0009]
  Claims of the invention5The invention described in claim 1 to claim 14The voice recognition device according to any one of the above, wherein the vehicle control command generation means generates a vehicle control command only when the voice input information is actually input simultaneously with the operation input information. Is a voice recognition device.
[0010]
  Claims of the invention6The invention described in claim 1Any one of 1 to 5In the voice recognition device according to the above, if the accuracy calculation unit has additional voice input information and / or operation input information after the accuracy is calculated, the accuracy is recalculated based on the additional information. Is a voice recognition device characterized by
[0011]
  Claims of the invention7The invention described in claim 11Or6In the voice recognition device according to any one of the above, a range in which the vehicle control command generation unit receives information on the current travel speed and acquires the road information according to the current travel speed, and to each branch The speech recognition apparatus is characterized in that a weight to be given to the transition accuracy is determined.
[0012]
  Claims of the invention8The invention described in claim 11Or7In the voice recognition device according to any one of the above, the vehicle control command generation unit includes:Road information (distance factor) according to the force applied to the operation input deviceIs a speech recognition apparatus characterized by determining the
[0013]
  Claims of the invention9The invention described in claim 1 to claim 186. The voice recognition device according to claim 1, wherein the vehicle control command generation unit further includes a control command presenting unit that presents information about the vehicle control command to a driver, and the vehicle control command is transmitted to the vehicle control unit. The voice recognition device is characterized in that the control command presenting means presents information on the vehicle control command when sending to the vehicle.
[0014]
  Claims of the invention10The invention described in claim 11Or9In the voice recognition device according to any one of the above, when the vehicle control command generation unit determines that the vehicle control command is not immediately sent to the vehicle control unit, the vehicle control command related to the transition to the branch with the highest accuracy Control command confirming means for presenting to the driver a question as to whether or not to execute the control command presenting means, and the vehicle control command generating means responds to voice input and / or operation input for the question. A speech recognition apparatus that determines whether or not to send a vehicle control command to the vehicle control means.
[0015]
  Claims of the invention11The invention described in claim 110In the voice recognition device according to claim 1, when the control command confirmation means confirms whether or not the vehicle control command is executed with respect to the driver, when the target in the vehicle control command approaches within a predetermined distance, The voice recognition device is characterized by sending a command to decelerate automatically to the vehicle control device.
[0016]
  Claims of the invention12The invention described in claim 110Or11If the response from the driver is not within a predetermined time, the vehicle control command is sent to the vehicle control device according to the branching accuracy corresponding to the vehicle control command, or the same It is a voice recognition device that selects whether to present a question again or to present a question to the driver regarding whether or not to execute a vehicle control command corresponding to a branch with the next highest accuracy.
[0017]
【The invention's effect】
  According to the first aspect of the present invention, voice input and operation input are input in combination as a command from the driver., And recognizing information on the road on which the vehicle is runningTherefore, the command intended by the driver can be estimated with high accuracy, so that the intended command can be transmitted to the vehicle control device easily and quickly.At the same time, the voice input information, the operation input information, and the road information are evaluated by calculating the accuracy representing each likelihood, so that accurate estimation can be performed..
[0018]
  Claim2According to the invention of the present invention, the vehicle control command can be issued even if not all of these are included in the voice input information by recognizing separately the vehicle motion target by voice command, the motion with respect to the target, and the direction of motion. Can be generated.
[0019]
  Claim3According to this invention, in order to determine whether or not to immediately execute the vehicle control, the accuracy representing the certainty of the information is calculated. Therefore, when the accuracy is low, it is possible to change the correspondence so that there is no sense of incongruity. become.
[0020]
  Claim4According to the invention, since the control command can be generated within the range of the obtained information in the command input from the driver and the road information, a flexible command suitable for the driver's feeling is possible.
[0021]
  Claim5According to the invention, since it is explicitly confirmed that the driver gives a voice command to the vehicle by the operation input, it is possible to deal with unintentional voice utterances such as radio voice, conversation with a passenger in the passenger seat, and monologue. It becomes possible to prevent malfunction.
[0022]
  Claim6According to the invention, when the accuracy is lower than the predetermined value, it is possible to quickly give information related to the control command to be performed on the vehicle without any sense of incongruity by additionally inputting only the command as the point.
[0023]
  Claim7According to the invention, since the range of the branch information to be searched is adjusted according to the traveling speed, it is possible to improve the voice recognition accuracy by obtaining the collation with the voice command within the range.
[0024]
  Claim8According to the invention, since it becomes easy for the driver to issue the control command quickly, the vehicle operation shifts to the branch near the own vehicle with respect to the force applied unconsciously or consciously. Generation is possible.
[0025]
  Claim9According to the invention, since the command and a series of operations executed by the vehicle are presented to the driver and the operations performed by the vehicle are transmitted to the driver, it is possible to improve the sense of security.
[0026]
  Claim10According to the invention, since a highly probable command is presented without re-inputting, the burden on the driver can be reduced.
[0027]
  Claim11According to the invention, when a command object having a high possibility of branching is approaching, the vehicle decelerates so that the vehicle does not pass therethrough, so that the intended command is passed while checking the vehicle. It becomes possible to prevent that.
[0028]
  Claim12According to this invention, it is possible to configure a command system that does not have a sense of incongruity that matches the driver's feeling by changing the correspondence according to the accuracy.
[0029]
DETAILED DESCRIPTION OF THE INVENTION
  FIG. 1 is a block diagram showing an example of a vehicle control system including an embodiment of a speech recognition apparatus according to the present invention. This system, for example, controls the traveling of an autonomous vehicle that can autonomously control the accelerator, brakes, and steering, and that can accept commands from the driver. It includes a recognition device 2, a driver interface unit 4, a travel position detection unit 6, a vehicle control unit 8, and a communication unit 10. The driver interface unit 4 includes, for example, a voice input device 12 that supplies a voice input command from the driver to the voice recognition device 2, a display device 14 that presents vehicle transition operation information from the voice recognition device 2 to the driver, An operation input device 16 that supplies an operation input command from the driver to the recognition device 2 and a voice utterance device 18 that presents an operation confirmation from the voice recognition device 2 to the driver are provided. The traveling position detection unit 6 includes, for example, a GPS navigation device 20 that supplies vehicle position information to the voice recognition device 2 and a magnetic nail detection device 22. The vehicle control unit 8 receives, for example, vehicle control information from the voice recognition device 2 and controls an automatic steering device 24 for controlling the steering of the vehicle, a radar device 26 for maintaining a distance between the front vehicle, and a voice recognition device. 2 is provided with a vehicle speed control device 28 for receiving the vehicle control information from 2 and controlling the speed of the vehicle. The communication unit 10 includes, for example, a road-vehicle communication device 30 and a vehicle-vehicle communication device 32.
[0030]
  One embodiment of the speech recognition apparatus 2 according to the present invention shown in FIG. 1 includes speech input recognition means 34, operation input recognition means 36, weight determination means 38, road information recognition means 40, accuracy calculation means 42, The vehicle control command generation means 44, the control command presentation means 46, the control command confirmation means 48, and the vehicle control means 50 are provided.
[0031]
  FIG. 2 is a flowchart for explaining the operation of the speech recognition apparatus 2 in FIG. In step S <b> 1, the voice recognition unit 34 receives voice input information from the voice input device 18 of the driver interface 4. Thereby, the speech recognition apparatus 2 starts a speech recognition operation.
[0032]
  In step S <b> 2, the operation input recognition unit 36 receives operation input information from the operation input device 14 of the driver interface 4. Note that the voice recognition operation in step S1 may be started by the operation input information in step S2, and if there is no operation input for a while after the voice is detected, the voice recognition operation may be stopped. Moreover, you may recognize the audio | voice detection result before and behind the operation input.
[0033]
  In step S3, the road information recognition means 40 receives the vehicle position information determined by the GPS navigation device 20 and the magnetic nail detection device 22 of the traveling position detection unit 6 and road information including attributes. The road information includes communication from the infrastructure and the dedicated radio station, communication with surrounding vehicles, or an external recognition sensor of the own vehicle by the road-vehicle communication device 30 or the vehicle-vehicle communication device 32 of the communication unit 10. For example, information such as a map of a certain range around the own vehicle shown in FIG. 3 and the positions of the surrounding vehicles is shown. It should be noted that lane change prohibition sections and branching point information determined by traffic regulations are explicitly shown as road attributes, for example using data as shown in FIG. 4 or map information supplemented by communication means Also good. Further, as the road attributes, a road width, the number of road lanes, a speed limit, and the like may be acquired.
[0034]
  In step S4, the road information recognition means 40 recognizes a route branch that can change the direction of the vehicle action as a branch that the vehicle can move from, based on the vehicle position information and road information obtained in step S3. , Generate branch information. The branch that can be transferred includes a road that can be transferred on the same road, such as a road branch such as a right or left turn, a vehicle change, a stop on the side of the road, or acceleration / deceleration. FIG. 5 schematically shows such branch information. Moreover, you may also hold | maintain the branch of continuing the present movement, without changing vehicle action.
[0035]
  In step S5, the voice input recognition unit 34 extracts a command target from the voice input information by collating with dictionary information. The command target includes a place name (for example, “Yokohama”), a turning point (for example, “next intersection”), a facility (for example, “gas station”), and the like. When collating with words included in the dictionary information, all words having a likelihood equal to or higher than a certain threshold are extracted. As a recognition result, the relative accuracy of the recognized word is determined by the likelihood, for example, as to the certainty.
[0036]
  In step S6, the voice input recognizing means 34 extracts a command action from the voice input information by collation with dictionary information. The command operation includes “turn”, “enter”, “accelerate”, and the like. As in step S5, when collating with words included in the dictionary information, all words having a likelihood equal to or higher than a certain threshold are extracted. As a recognition result, the relative accuracy of the recognized word is determined by the likelihood, for example, as to the certainty.
[0037]
  In step S7, the voice input recognizing means 34 extracts the command direction from the voice input information by collation with dictionary information. The command directions include, for example, “right”, “left”, “right diagonal”, and the like. Again, as in steps S5 and S6, when collating with words included in the dictionary information, all words with a likelihood equal to or greater than a certain threshold are extracted. As a recognition result, the relative accuracy of the recognized word is determined by the likelihood, for example, as to how certain it is.
[0038]
  In addition, when selecting a word from dictionary information in step S5, S6, and S7, you may make it select only some words with the highest likelihood. Further, regarding the dictionary information used in steps S5, S6, and S7, a dictionary to be used may be set based on the dictionary information that is linked to the vehicle position. Further, it is not necessary to provide means for recognizing all of steps S5, S6 and S7. Alternatively, a command such as “turn left” may be extracted by separating it into “left” and “turn”, or a command such as “speed 40” may be extracted as converted to “acceleration” by comparison with the current vehicle speed. Good. At this time, for the portion corresponding to the control amount such as “40”, the accuracy of the control amount including the degree of coincidence with other numerical information such as “50”, “60”, and “20” is determined based on the likelihood. The information stored may be used when the vehicle control command is actually issued in step S13 described later.
[0039]
  In step S8, the road information recognizing means 40 determines, for example, the ease of travel of the vehicle related to the structure of the road that can be acquired from the map information mounted on the vehicle, such as the capacity determined by the width of the road and the speed limit. Information is acquired as road structure information.
[0040]
  In step S9, the road information recognition means 40 acquires information determined by the vehicle traveling. Such information includes, for example, a coefficient for ease of travel as a ratio of the traffic speed to the speed limit. For example, the speed of traffic may be estimated from the speed of a preceding vehicle measured by an external sensor, or may be obtained from road-to-vehicle communication during traveling. In addition, when a route for traveling in advance is set by the navigation system, the weight of the set route may be set to be large, for example, the weight is doubled.
[0041]
  In step S10, the weight determination means 38 determines a distance coefficient. This distance coefficient is a weighting coefficient regarding the probability of shifting to each branch obtained in step S4, and a to d in FIG. 6 are graphs explaining such a distance coefficient. For example, as shown in FIG. 6a, the weight is defined so as to change according to the distance from the own vehicle to the command target. When the force applied to the operation input device or the amount of change in operation increases, the weight near the host vehicle may increase as shown in FIG. 6b. Further, such weighting characteristics may be changed according to the traveling speed of the host vehicle. For example, when the traveling speed is high, it is difficult to physically change the vehicle route in the vicinity of the host vehicle as shown in FIG. The weight may be set to zero, or the front distance to which the weight is applied may be extended. Further, as shown in FIG. 6d, the weight may be set to change exponentially.
[0042]
  In step S11, the accuracy calculation means 42 uses the information obtained in steps S2, S5, S6, S7, S8 and S9 and the distance coefficient determined in step S10 to determine the accuracy of transition to each branch obtained in step S4. And is supplied to the vehicle control command generation means 44. This calculation method will be described in detail later.
[0043]
  In step S12, the vehicle control command generation unit 44 determines whether or not to immediately supply the vehicle control command for shifting to the branch having the highest accuracy calculated in step S11 to the vehicle control unit 50 according to the accuracy. To do. Here, as an example, when the accuracy value is 0.8 or more, it is extremely high, when it is 0.65 or more and less than 0.8, it is high, and when it is 0.5 or more and less than 0.65, it is medium or 0.5 or less. In this case, it is specified as low. If it is extremely high, the process proceeds to step S13; otherwise, the process proceeds to step S14.
[0044]
  In step S13, the corresponding vehicle control command is immediately supplied to the vehicle control means 50. At the same time, the control command presenting means 46 may present the driver to execute this vehicle control command through the display device 14 and / or the voice utterance device 18 of the driver interface unit 4. Furthermore, when a route for traveling in advance is set by the navigation system 20, the route may be rerouted to the selected branch. Upon receiving the vehicle control command, the vehicle control means 50 controls the vehicle control unit 8 based on the vehicle control command while taking into account information related to surrounding vehicles and causes the vehicle to travel according to the command. The time from when the information is presented to the driver until when the command is actually started may be changed according to the accuracy. In addition, when a cancel command for canceling the command is input within a time during which the vehicle control command is actually supplied to the vehicle control means 50 and can be stopped even after the vehicle control command is supplied, the command is canceled. It is also possible to jump to step S14 instead. In this case, when a continuous cancel command is input, it may be completely canceled. Also, until the vehicle control command is actually supplied to the vehicle control means 50, voice input and / or operation input is received even during the time when it can be stopped even after being supplied, and these inputs are taken into consideration. The calculated accuracy may be recalculated, and the process may return to step S12 based on the result. Further, as described above with respect to steps S5, S6, and S7, processing is performed in the case where the control amount is designated as the speed “40” by the voice command, which will be described later.
[0045]
  In step S22, after the vehicle control command is executed, the held accuracy to each branch and the vehicle control command are reset, and the system waits for the next voice input and / or operation input.
[0046]
  In step S14, the vehicle control command corresponding to the branch having the highest accuracy is selected from the remaining branches excluding the branch already denied by the driver.
[0047]
  Steps S15, S20, and S21 are routines for determining how close the command target of the vehicle control command selected in step S14 is to prevent overrun.
[0048]
  In step S15, the distance to the command target of the vehicle control command selected in step 14 is determined, and the speed that needs to be decelerated when the command target is reached in consideration of the current vehicle speed and acceleration. It is determined whether or not a certain degree of deceleration is necessary to reach the target. If it is determined that the deceleration is necessary, the process proceeds to step S20. Otherwise, the process proceeds to step S21.
[0049]
  In step S20, the vehicle control means 50 controls the vehicle control unit 8 to start deceleration so as not to overrun the branch point corresponding to the command target, and to the driver by the display device 14 and / or the voice utterance device 18. Encourage them to switch to manual operation. If the accuracy corresponding to the extreme height in step S12 is obtained by adding an additional input in the meantime, the process proceeds to step S13 in order to supply the corresponding vehicle control command to the vehicle control means 50. But you can. Note that steps S15, S20 and S21 may be omitted.
[0050]
  In step 21, it is determined whether or not switching to manual operation has been performed. When the switching to the manual operation is performed, the automatic steering 24 and the vehicle control device 28 are released. If not switched to manual operation, the process proceeds to step S16.
[0051]
  In step S16, the control command confirmation unit 48 inquires of the driver whether or not to execute the vehicle control command selected in step S14 by the display device 14 and / or the voice utterance device 18.
[0052]
  In step S17, it is determined whether or not there is a response from the driver. If there is a response within a predetermined time, the process proceeds to step S18, and if there is no response, the process proceeds to step S19. Such a reply may be by voice or by operation.
[0053]
  In step S18, it is determined whether or not the reply indicates affirmation. If yes, the process proceeds to step S13. Otherwise, the process proceeds to step S14, and the same processing is performed regarding the vehicle control command relating to the branch with the next highest accuracy. If the reply is neither affirmative nor negative, it is possible to jump to step S16 and confirm again with the driver.
[0054]
  In step S19, it is determined whether the accuracy of the branch currently being processed is high, medium, or low. If high, the process proceeds to step S13, if medium, the process proceeds to step S16, and if low, the process proceeds to step S14. Proceed with Note that processing branches performed according to the accuracy in step S19, such as a mode that proceeds to step S14 even if medium or low, and a mode that proceeds to step S16 even if high or medium, are individually performed in step S13. , S14, and S16 may be set to be connected. However, as a result of assigning the accuracy in the order of steps S13, S16, and S14, the accuracy to S16 is set to S14 or higher, and the accuracy to S13 is set to be S16 or higher. Further, in step S14, for example, all the held commands may be cleared by canceling the command at an arbitrary time point by operating the command cancel switch or the like.
[0055]
  The accuracy calculation method in step S11 described above will be described. As an example for explanation, while driving on a road as shown in FIG. 3, in order to make a left turn in the direction of Yokohama at the previous intersection, say “turn to Yokohama” and turn left on the operation input device. Consider the case of operation. That is, it is assumed that road information as shown in FIG. 3 is recognized at the time of such an operation, and branch information as shown in FIG. 5 is obtained.
[0056]
  FIG. 7 is a diagram for explaining the process of calculating the accuracy in the case of such an example. In the flow F1, as a result of the voice recognition for the voice “turn to Yokohama”, the items related to the command target included in the branch information in this example are 0.6 for “Yokohama”, 0.3 for “Yokosuka”, “Gas station” ”Is recognized, and items relating to the command action are recognized as 0.6 for“ turn ”, 0.2 for“ enter ”, and 0.2 for“ acceleration ”.
[0057]
  In the flow F2, a probability of shifting to each branch in the branch information is assigned. Weighting is performed according to the likelihood recognized as described above. In this example, a value of 0.6 is given to the branch corresponding to the command target “Yokohama”, 0.3 to the branch corresponding to “Yokosuka”, and 0.1 to the branch corresponding to “gas station”. A certain value, 0.1 in this example, is given to a branch corresponding to an item that has no likelihood equal to or greater than a preset threshold value. Similarly, 0.6 corresponds to the branch corresponding to “turn” of the command operation, 0.2 corresponds to the branch corresponding to “change lane”, 0.2 corresponds to the branch corresponding to “enter”, and corresponds to “acceleration”. A value of 0.2 is given to the branch, and a constant value, 0.1 in this example, is given to the branch corresponding to an item that has no likelihood equal to or greater than a preset threshold value.
[0058]
  In this example, there is no voice input related to the command direction, but in some cases, a weight value is given to each branch in the same manner as the other command items. When there is no related voice input as in this example, a constant value may be given to all branches. The same applies when there is no voice input related to other commands.
[0059]
  The accuracy related to the operation input information is set so that the highest value is given to the branch corresponding to the direction input in the operation input device, and the value given to the other branch is set so that the weight becomes lighter as it deviates from this direction. . When there are a plurality of recognized directions, a plurality of weights may be overlapped. In addition, for example, when the intersection of the command target is a three-dimensional intersection, the operation input is put in the right direction when it is necessary to branch leftward from the currently running road in order to go right. However, it may be considered that the input has been made in the left direction, and the weight corresponding to the branch corresponding to the left direction may be increased.
[0060]
  As for the accuracy related to the road structure information, the weighting value of each route is determined according to the road capacity or the like.
[0061]
  As for the accuracy related to the road condition information, the weight value of each route is determined according to the ease of driving of the vehicle such as the degree of congestion and the current traveling speed.
[0062]
  Further, a distance coefficient as described with respect to step S10 in FIG. 2 is given to each path.
[0063]
  In the flow F3, the accuracy regarding each information given in the flow F2 is multiplied for each branch. In the case of this example, the weights for multiplying the accuracy related to each information are made equal, but the weights may be different among the information.
[0064]
  The flow F4 is the finally calculated accuracy, and the numerical value is such that the sum of the transition accuracy for all branches is 1, the accuracy σ = (the highest branch accuracy σ_i) / (Σ accuracy of each branch σ_n) Is standardized by calculation. From this result, it can be seen that the branch with the highest accuracy is the branch corresponding to the left turn in the direction of Yokohama, and the accuracy is 0.86.
[0065]
  FIG. 8 is a diagram for explaining a process of calculating the same accuracy as that in FIG. 7, but in the example shown in this diagram, the accuracy for each information is multiplied step by step. Flows F1 and F2 are the same as those in FIG. 7, but in flow F3, the accuracy obtained by multiplying only the weight related to the voice input information, the accuracy obtained by multiplying only the weight related to the operation input information, the road information, and the distance The accuracy obtained by multiplying only the weights related to the coefficients is shown. At this stage, the branching accuracy corresponding to the left turn toward Yokohama is 0.54, 0.28, and 0.11, respectively.
[0066]
  In the next flow F3 ', the accuracy is shown when two of the three accuracy values calculated in the flow F3 are extracted and combined. At this stage, the branching accuracy corresponding to the left turn toward Yokohama is 0.59 (voice input information + operation input information), 0.69 (voice input information + road information and distance coefficient), 0.48 ( Operation input information + road information and distance coefficient).
[0067]
  When the values calculated in the flow F3 are multiplied in the flow F4, the branching accuracy corresponding to the operation of turning left in the direction of Yokohama is 0.86, which is the same as in the case of FIG. Therefore, the accuracy of the intermediate stage such as the flows F3 and F3 ′ in this figure is not necessarily required to calculate the final accuracy, but due to the property that the accuracy of each information can be superimposed in this way, the intermediate stage Or, once the final accuracy is calculated, if there is additional input related to certain information, the accuracy of each additional input information is multiplied by the accuracy calculated at that time to obtain the accuracy of each branch. A new calculation can be performed.
[0068]
  FIG. 9 is a diagram similar to FIG. 7, but illustrates the accuracy calculation process when “lane change” is performed as a voice input and a rightward operation is performed as an operation input. The accuracy of is assumed to be 0.7 for “lane change”, 0.2 for “turn”, and 0.2 for “acceleration”. The value calculated in the same manner as in FIG. 7 is 0.96, which is the highest branch for “lane change” here.
[0069]
  A process when the control amount is designated like the speed “40” by the voice command in steps S5, S6, and S7 in step S13 described above will be described. In such an example of speed, the current vehicle speed information is compared with road speed limit information, etc., and the accuracy of reaching the speed limit is the highest, for example, as in the calculation of the accuracy in step S11, and decreases with increasing distance. And the control amount may be calculated by multiplying the likelihood of the speech recognition result as shown in FIG.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an example of a vehicle control system including an embodiment of a voice recognition device according to the present invention.
FIG. 2 is a flowchart illustrating the operation of an embodiment of a speech recognition apparatus according to the present invention.
FIG. 3 is a diagram of map information representing road information.
FIG. 4 is a diagram of map information supplemented with road attributes.
FIGS. 5a and 5b are diagrams schematically showing branch information. FIGS.
FIGS. 6A to 6D are graphs for explaining a distance coefficient.
FIG. 7 is a diagram illustrating a process of calculating accuracy.
FIG. 8 is a diagram illustrating a process for calculating accuracy.
FIG. 9 is a diagram illustrating a process for calculating accuracy.
FIG. 10 is a diagram illustrating processing when a control amount is specified.
[Explanation of symbols]
  2 Voice recognition device
  4 Driver interface section
  6 Traveling position detector
  8 Vehicle control unit
  10 Communication Department
  12 Voice input device
  14 Display device
  16 Operation input device
  18 Voice utterance device
  20 GPS navigation device
  22 Magnetic nail detector
  24 Automatic steering device
  26 Radar equipment
  28 Vehicle speed control device
  30 Road-to-vehicle communication device
  32 Inter-vehicle communication device
  34 Voice input recognition means
  36 Operation input recognition means
  38 Weight determining means
  40 Road information recognition means
  42 Accuracy calculation means
  44 Vehicle control command generation means
  46 Control command presentation means
  48 Control command confirmation means
  50 Vehicle control means

Claims (12)

Voice input recognition means for recognizing voice input information;
Operation input recognition means for recognizing operation input information;
And a vehicle control command generation means for generating a vehicle control command based on the command and the voice input information and the operation input recognizing means for said speech input recognition means recognizes recognizes,
Vehicle control means for controlling acceleration / deceleration and steering of the vehicle based on the vehicle control command generated by the vehicle control command generation means ;
The branch information to which the vehicle can move, the road structure such as the lane width obtained from the map information, the road condition such as the congestion degree obtained from the external recognition sensor and the infrastructure information, and the distance to each branch of the branch information Road information recognition means for recognizing road information including at least one of distance coefficients that are weights according to
The vehicle control command generating means, comprising: an accuracy calculating means for calculating a probability that the vehicle in the branch information is transferred to each branch to which the vehicle can move based on the voice input information, operation input information, and road information. but the speech recognition apparatus characterized that you generate a command to perform a shift to the highest probability of the branch as the vehicle control command.   2. The voice recognition device according to claim 1, wherein the voice input recognition means recognizes at least one of a target of a vehicle control action, an action with respect to the target, and a direction of the action in the voice input information. A speech recognition apparatus characterized by that. 3. The voice recognition apparatus according to claim 1 , wherein the vehicle control command generation unit determines whether or not to immediately send the vehicle control command to the vehicle control unit according to the accuracy. Recognition device. 4. The voice recognition device according to claim 1 , wherein the accuracy calculation unit calculates the accuracy of shifting to each branch to which the vehicle can shift, and the voice input information and the operation input. 5. A speech recognition apparatus characterized by calculating the accuracy based on the remaining information even if there is missing information in the information and road information. The speech recognition apparatus according to any one of claims 1 to 4, wherein the vehicle control command generation means, a vehicle control command only when the speech input information is input at the same time in practice and the operation input information A speech recognition apparatus characterized by generating. 6. The voice recognition device according to claim 1 , wherein when the accuracy calculation means has additional voice input information and / or operation input information after the calculation of the accuracy, these additional information is provided. A speech recognition device characterized by recalculating the accuracy based on the above. The voice recognition device according to any one of claims 1 to 6 , wherein the vehicle control command generation means receives information on the current travel speed and acquires the road information according to the current travel speed. And a weight to be given to the transition accuracy to each branch. The speech recognition apparatus according to any one of claims 1 to 7, said vehicle control command generation means, to determine the road information (distance factor) according to the magnitude of the force applied to the operation input device A voice recognition device characterized by the above. The speech recognition apparatus according to any one of claims 1 to 8, wherein the vehicle control command generating means further comprises a control command presenting means for presenting information about the vehicle control command to the driver, the vehicle control command The voice recognition apparatus, wherein when the control command presenting means is sent to the vehicle control means, the control command presenting means presents information on the vehicle control command. The speech recognition apparatus according to any one of claims 1 to 9 , wherein when the vehicle control command generation unit determines that the vehicle control command is not immediately sent to the vehicle control unit, the branch to the branch with the highest accuracy is performed. Control command confirmation means for presenting to the driver a question as to whether or not to execute a vehicle control command related to the transition is provided to the driver by the control command presenting means, and the vehicle control is performed according to a voice input and / or an operation input for the question A voice recognition device, wherein the command generation means determines whether or not to send the vehicle control command to the vehicle control means. 11. The voice recognition device according to claim 10 , wherein when the control command confirmation unit confirms whether or not the vehicle control command is executed with respect to the driver, a target in the vehicle control command approaches within a predetermined distance. In this case, the voice recognition device is characterized by sending a command to automatically decelerate to the vehicle control device. 12. The voice recognition device according to claim 10 or 11 , wherein when the response from the driver is not within a predetermined time, the vehicle control command is sent to the vehicle control device according to the branching accuracy corresponding to the vehicle control command. Or presenting the same question again, or presenting the driver with a question as to whether to execute a vehicle control command corresponding to the next most likely branch. Recognition device.

Images