JP2006317573A - Information terminal - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an information terminal that can acquire information for a user through voice interaction with the user. <P>SOLUTION: The information terminal which recognizes user's voice as a command and provides information, based on the recognized command comprises a talk signal part for outputting a talk signal, according to a user's instruction; a voice recognition part for recognizing the voice uttered by the user as the command; an environmental information acquiring part for acquiring environmental information about the environment in the surroundings of the information terminal; an inferring part for inferring the information intended by the user from the command recognized and the environmental information acquired; and a control part for controlling the processing of the information terminal, and on receiving the talk signal, the control part for deciding whether the user's command, recognized by the speech recognition part, has been acquired, and provides the user with the information presumed by the inferring part for using the environmental information acquired by the environmental information acquiring part, when the command has not been obtained. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an information terminal that obtains information from a user through voice interaction with the user.

  In-vehicle information terminals such as current car navigation systems hold a large amount of information such as store information and facility information in a large-capacity storage medium such as a DVD or HDD. Furthermore, these pieces of information can be acquired from the server using wireless communication.

  In a general information retrieval apparatus, a method is used in which a user performs an operation of narrowing down necessary information from a large amount of information by inputting and selecting a keyword, a classification name, or the like. Even in a vehicle information terminal, since this method is effective while the vehicle is stopped, an information search interface using a keyword input by a user using a software keyboard and a method of selecting a classification on a menu screen is adopted. On the other hand, a system has been developed that uses a voice recognition system to extract necessary information hands-free / eyes-free when a user inputs a keyword or a classification name by voice.

  In addition, unlike a general web search engine, information required by a user may be greatly influenced by surrounding conditions such as where the user is running and whether a traffic jam occurs. Therefore, there is also known a method for improving the convenience of the user by providing information only in the vicinity where the user's vehicle is traveling or the store information and facility information on the route scheduled to travel ( (See Patent Document 1).

Further, a Bayesian network can be used as a method for estimating information based on a user's voice command. A Bayesian network is a probabilistic model with a graph succession, and can be used for predicting events including uncertainty, rational decision making, and fault diagnosis to find the cause from observation results (Non-Patent Documents) 1).
JP-A-11-51666 Yoichi Motomura, “Bayesian Network: From Introduction to Application to Human Modeling”, [online], National Institute of Advanced Industrial Science and Technology, [Search April 1, 2005], Internet <URL: http: //staff.aist. go.jp/y.motomura/paper/BSJ0403.pdf>

  However, even if the accuracy of estimation is improved by improving the method for estimating information required by the user, it is difficult to accurately estimate the timing when the user wants to know the information.

  For example, when a traffic jam occurs, it can be said that the traffic jam information is appropriate as an estimated value of the type of information that the user needs in the situation. However, if the user already knows the traffic jam information by another means, or if the user is prepared for the traffic jam from the beginning, it is not pleasant for the user to be informed of the traffic jam distance and time.

  In addition, it is possible to provide a service for providing nearby store information based on vehicle position information and user purchase history. In this case, the store providing information wants to provide information as frequently as possible, but the user may feel troublesome that such information is frequently provided.

  In addition, there is a method to provide information to the user by estimating information to be provided to the user and operations to be performed periodically, but when it is estimated that information in a certain situation is very likely to be required However, since there may actually be a mismatch that the user does not want information at that timing, this may similarly be annoying for the user to be provided with information.

  Because of these problems, push-type information provision from information terminals is currently not widespread for in-vehicle information terminals, and a method of specifying information provided by user operation as described in the prior art is generally used. Is.

  The problem here is that when speech recognition is used as an input means, the user cannot think of an appropriate voice command and speaks nothing, the wrong voice command is spoken, or the ambient noise No recognition result is obtained at all, such as when the recognition process is mistaken for user utterance. In such a situation, the conventional system merely urges the user to recite the voice, and the user utters the voice many times.

  The present invention recognizes a user's voice as a command, and an information terminal that provides information based on the recognized command, recognizes a talk signal unit that outputs a talk signal according to a user's instruction, and recognizes a voice uttered by the user as a command. A voice recognition unit, an environment information acquisition unit that acquires environment information about the environment around the information terminal, an estimation unit that estimates information intended by the user from the recognized command and the acquired environment information, and processing of the information terminal A control unit that controls, the control unit determines whether or not the user's command recognized by the voice recognition unit is obtained in response to reception of the talk signal. The information estimated by the estimation unit using the environment information acquired by the information acquisition unit is provided to the user.

  According to the present invention, even when the user does not come up with an appropriate voice command and does not utter anything, or when the system utters unexpectedly, the user's operation is not wasted and the situation is not affected. Accordingly, it is possible to provide information that is likely to be required by the user. Thereby, a highly convenient user interface can be provided.

  Furthermore, since the user intention is estimated only when the user presses the talk button, there is an advantage that the user is not pressed against unnecessary information.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing the configuration of an information terminal according to the first embodiment of the present invention.

  The information terminal 100 is installed in a vehicle such as an automobile. Then, an instruction by a voice command by the user's voice is received, the voice command is recognized, and information corresponding to the recognition result is responded by a synthesized voice or display on a display.

  The information terminal 100 includes a system control unit 101, a GPS receiver 102, an autonomous navigation sensor 103, a traffic information receiver 104, a display 105, an audio output device 106, a hard disk (HDD) 109, a microphone 110, a talk button 111, and a GUI operation. Unit 112, AV device 113, road surface sensor 116, sweat sensor 117, memory 118, interface 119, and the like.

  The system control unit 101 executes processing of the information terminal 100. The system control unit 101 includes a CPU, a ROM, a RAM, a data bus, and the like.

  The GPS receiver 102 is a device that receives signals from a plurality of GPS satellites (not shown) and calculates the current position of the vehicle.

  The autonomous navigation sensor 103 includes a gyro and is a device that estimates the current position of the vehicle and the direction of the vehicle using information from the gyro and vehicle speed pulses acquired from the vehicle via the interface 119.

  The traffic information receiver 104 is a device that receives a traffic information signal carried by a beacon or FM broadcast wave.

  The display 105 is a device that displays map information, vehicle position information, traffic information, and the like to the user. The display 105 is configured by a liquid crystal display device, for example.

  The voice output device 106 responds to the user with the synthesized voice in response to an instruction from the system control unit 101. For example, route guidance while traveling is transmitted to the user by synthetic voice.

  A hard disk device (HDD) 109 is a storage medium for storing digital data. The HDD 109 stores map information.

  The microphone 110 is a device for collecting voice commands uttered by the user. A plurality of the microphones 110 are provided in the vehicle, and it can be determined from the difference in the arrival times of the voices acquired by the plurality of microphones whether the voice is produced by the driver seat user or the passenger seat user.

  The talk button 111 generates a talk signal that instructs the voice recognition unit 114 to execute voice recognition processing. Specifically, after the user presses the talk button 111, the voice command is uttered. The system control unit 101 causes the voice recognition unit 114 to execute voice command recognition processing when a talk signal is received by pressing the talk button. Accordingly, it is possible to know the timing at which the user issues a voice command. For example, it is possible to prevent erroneous recognition due to noise even in a running noise environment and to obtain good voice recognition performance.

  The talk button 111 is configured by a switch attached to a vehicle handle, for example. Any means for generating a talk signal for notifying the system control unit 101 that the user speaks may be used.

  For example, the voice button may be used instead of a talk button by uttering a specific keyword before uttering a voice command. In this case, the voice recognition unit 114 always waits for only a specific keyword (specific vocabulary word spot), and transmits a talk signal to the system control unit 101 when the voice recognition unit 114 detects the specific keyword. That is, the user speaking a specific keyword corresponds to pressing the talk button 111.

  The GUI operation unit 112 is a device for a user to operate the information terminal by a method other than a voice command, and includes a remote controller, a touch panel, buttons, and the like.

  The AV device 113 includes a CD player, an MD player, a DVD player, a composite player of these, a radio tuner, a television tuner, an amplifier, and the like. As these playback results, audio is output from the audio output device 106, and video is displayed on the display 105.

  The road surface sensor 116 is a sensor that detects the state of the road surface during travel (dry, wet, water film, snow cover, freezing). The road surface sensor 116 is, for example, a visible image sensor, a laser radar sensor, or an optical fiber sensor. A sensor or the like is used.

  The perspiration sensor 117 is attached to the handle, detects perspiration of the user during driving, and measures the user's tension.

  The interface (I / F) 119 is connected to various sensors of the vehicle and acquires various information such as a vehicle speed pulse, a handbrake signal, and an engine start signal.

  The memory 118 stores a route search unit 107, a map information reading unit 108, a voice recognition unit 114, and an intention estimation unit 115. Each of these units is a program executed by the system control unit 101.

  The route search unit 107 searches the map information stored in the HDD 109 for a route to the destination set by a user instruction or the like.

  The map information reading unit 108 reads the map information stored in the HDD 109 into the memory 118. Then, map matching is executed using the position information calculated from the information received by the GPS receiver 102, the position information obtained by the autonomous navigation sensor 103, and the read map information, and the current position of the vehicle is estimated. Then, a map around the estimated vehicle position and a symbol representing the vehicle are displayed on the display 105.

  The voice recognition unit 114 obtains a voice command collected by the microphone 110 as a recognition result of text or a symbol corresponding to the voice command. More specifically, the similarity between the voice model held by the voice recognition unit 114 and the input voice command is calculated, and the one with the higher similarity is used as the recognition result. Then, the recognition result and its similarity are output.

  The intention estimation unit 115 receives the position information calculated by the GPS receiver 102 and the autonomous navigation sensor 103, the traffic situation received by the traffic information receiver 104, the road condition detected by the road sensor 116, and received from the interface 119. Information required by the user or desired operation content is estimated from vehicle information and the like.

  FIG. 2 is a flowchart of the voice interaction process of the information terminal 101 according to the first embodiment of this invention.

  This flowchart shows a process of how the information terminal 100 recognizes the voice command and responds when the user presses the talk button 111 to utter a voice command.

  First, the user presses the talk button 111 to notify the information terminal 100 of the utterance of the voice command. When the talk button 111 is pressed, a talk signal is transmitted to the system control unit 101.

  When the system control unit 101 receives the talk signal from the talk button 111 (S1001), the system control unit 101 generates a question content for prompting the user to speak (S1002). At this time, that is, the question content generated immediately after the talk button 111 is pressed is, for example, a short voice such as “What?” Or “Please speak” or a short signal sound such as “Pip”. ing.

  The system control unit 101 instructs the voice output device 106 about the generated question content. The voice output device 106 outputs the contents of the question as voice (S1003). At this time, the content of the question may be displayed as text on the display 105 at the same time.

  In response to the output of the question content, the user utters a voice command (S1004). The microphone 110 collects voice commands that have been uttered and performs A / D conversion. The A / D converted voice command is sent to the voice recognition unit 114. The voice recognition unit 114 obtains a recognition result by performing voice recognition processing on the voice command (S1005).

  In addition, the result of this voice recognition processing is obtained when the text or symbol corresponding to the voice command uttered by the user is obtained as the voice recognition result and when the text or symbol corresponding to the user utterance is not obtained as the voice recognition result. There are two ways.

  In the latter case, for example, when the user presses the talk button 111 and does not utter anything for a certain time (timeout), the voice uttered by the user is a voice that the voice recognition unit 114 does not assume. Or when the user presses the talk button 111 and ambient noise is collected by the microphone 110, and the speech recognition unit 114 mistakes this noise for the user's speech and performs speech recognition processing. In such a case, the similarity between the speech model held by the speech recognition unit 114 and the input speech is low, and a recognition result cannot be obtained.

  The system control unit 101 determines whether an audio processing result is obtained as a result of the processing of S1005 (S1006). When the voice recognition result is obtained, the system control unit 101 uses the information included in the recognition result to narrow down the information required by the user and the candidate for the desired operation (S1007). The case where the voice recognition result is not obtained will be described later.

  Next, the system control unit 101 acquires current environment information. The environmental information is information acquired from at least one of the GPS receiver 102, the autonomous navigation sensor 103, the traffic information receiver 104, the route search unit 107, the map information reading unit 108, the road surface sensor 116, the sweat sensor 117, and the like. (S1008). The environmental information is information related to the situation of the car and the driver, and is different from the information inputted by voice.

  Next, the intention estimation unit 115 estimates the information required by the user or the desired operation content based on the environment information acquired in S1008 within the range of the recognition result narrowed down in S1007 (S1009). A specific method of this estimation will be described later.

  As a result of this estimation, the intention estimation unit 115 obtains the estimation content and the reliability of the estimation. Then, the system control unit 101 determines whether or not the estimation reliability is equal to or higher than a predetermined threshold (S1010).

  If the reliability is higher than the threshold, it is determined that the estimated content is equal to the user's intention. Therefore, information related to the estimated content is provided to the user (S1011). Specifically, the system control unit 101 presents information related to the estimated content by display on the display 105 or voice output by the voice output device 106. Alternatively, the estimated operation content is executed, and the user is informed that the operation content has been executed by display on the display 105 or sound output by the sound output device 106.

  On the other hand, the case where the similarity is low and the recognition result is not obtained in S1006 will be described.

  In a conventional system using voice recognition, if a user utterance recognition result is not obtained, it is regarded as a voice recognition failure. Therefore, the question content which asks the user to speak again is generated and the question content is output.

  On the other hand, in the present invention, when the recognition result is not obtained, the environment information is acquired (S1008), and the information required by the user or the desired operation content is obtained based on the acquired environment information. Estimate (S1009).

  As described above, the situation where the result cannot be obtained is that (A) when the user presses the talk button 111 and nothing is uttered for a certain period of time, (B) when the user does not assume the system (C) When the user presses the talk button 111, ambient noise is collected by the microphone 110, and the speech recognition unit 114 mistakes this noise for the user's speech and performs speech recognition processing. Can be classified. From the viewpoint of the user, this corresponds to a case where an appropriate voice command is not conceived even though the talk button 111 is pressed, or a case where an attempt is made to utter a voice command but the user is disturbed by ambient noise. At this time, it is certain that the user presses the talk button 111 and needs some information or tries to perform some operation by voice.

  Therefore, as in the conventional system, when the user's intention is estimated, the user's convenience is increased because the user's intention is not requested again, rather than the user's intention to be uttered again when the recognition result is not obtained. In particular, in the case of (A) and (B), there is a possibility that the user cannot come up with an appropriate voice command even if the utterance is requested again, so that the recognition result cannot be obtained again. This impairs the convenience of the information terminal 100 and lowers user satisfaction.

  Therefore, the information estimated in S1009 is provided to the user (S1011). That is, the system control unit 101 presents information related to the estimated content by display on the display 105 or sound output by the sound output device 106. Alternatively, the estimated operation content is executed, and the user is informed that the operation content has been executed by display on the display 105 or sound output by the sound output device 106.

  At this time, the estimated content or the operation content output to the user is not necessarily correct. Therefore, it is necessary to ask the user for confirmation or provide an opportunity for correction.

  In particular, when the intention estimation unit 115 determines that the reliability is low in S1010, the information to be provided to the user and the operation to be performed cannot be appropriately estimated. Therefore, the system control unit 101 generates a question content that requests further information from the user (S1002). The generated question is output to the user through the display 105 and the audio output device 106 (S1003), and the user's utterance is prompted.

  If the system control unit 101 determines that the reliability is high in S1010, the system control unit 101 first provides information to the user (S1011). Then, the system control unit 101 generates a question content that prompts confirmation utterance or correction utterance (S1002). The generated question is output to the user through the display 105 and the audio output device 106 (S1003), and the user's utterance is prompted. The contents of the question at this time are, for example, contents such as “Is there anything else?” Or “Are you sure about this operation?”

  As a result, when the recognition result of the user's voice command is not obtained, the information terminal 100 acquires the environment information, estimates the information required by the user or the desired operation content, and presents it to the user. In particular, even when the estimated content is wrong, the user can correct the utterance without the need to press the talk button 111 again, and the convenience of the user is increased. Furthermore, for the user, when making a request to the information terminal 100, the user always presses the talk button 111 to issue a voice command. As a result, it is possible to intuitively understand that the consistency of the user interface is maintained, and the convenience of the user is enhanced.

  In addition, in the conventional method of periodically providing information and operations to the user, even if it is estimated that there is a high possibility that information in a certain situation is required, the user actually However, there could be a mismatch that did not want information at that time. On the other hand, in the present invention, when the user presses the talk button 111, the information is estimated after the user knows that the information is necessary. Therefore, at the timing when the user needs the information. Necessary information is provided, there is no troublesomeness, and convenience is enhanced.

  FIG. 3 is an explanatory diagram of an example of an interaction between the information terminal 100 of the present invention and the user.

  Note that “S” represents information presentation or question by voice output from the information terminal 100, and “U” represents utterance by the user.

  In the example of FIG. 3A, when the user first presses the talk button 111, a talk signal is transmitted to the system control unit 101 (S1001). In response to this, the system control unit 101 generates question contents for response (S1002) and outputs them (S1003). The question content at this time is a voice “What is it?” Prompting the user to speak.

  At this time, since the user did not come up with an appropriate voice command, after the talk button 111 is pressed, the silence has been silenced for a predetermined time, and the voice recognition unit 114 determines that a timeout has occurred. Therefore, a voice recognition result cannot be obtained (S1006). Therefore, environmental information is acquired (S1008). In the example of FIG. 3, the traffic information receiver 104 acquires traffic jam information. Further, the position information obtained by matching the position information of the GPS receiver 102, the position information of the autonomous navigation sensor 103, and the map information obtained by the map information reading unit 108 proves that the user is on a congested road. is doing.

  Based on this environmental information, the intention estimation unit 115 estimates that the information required by the user is traffic jam information. And it determines with the reliability of this presumed content being higher than a predetermined threshold value.

  Therefore, the system control unit 101 provides traffic information to the user as shown in FIG. After that, a question content ("Is there anything else?") That prompts the user to confirm or correct the utterance is output.

  Since the information required when the talk button 111 is pressed is the traffic jam information, the user has obtained the information, and therefore responds “No” and ends the dialogue with the information terminal 100.

  As in this example, even when the user does not come up with an appropriate voice command, the user can obtain the necessary information.

  The example of FIG. 3B is a case where a voice recognition result cannot be obtained because the user did not come up with an appropriate voice command, as in the example of FIG. The intention estimation unit 115 estimates that information required by the user is traffic jam information. The system control unit 101 provides this traffic jam information to the user. After that, a question content ("Is there anything else?") That prompts the user to confirm or correct the utterance is output.

  However, the user recognizes that the information required when the talk button 111 is pressed is different from the provided information. Therefore, the “return information” which is necessary information is returned.

  In response to this, the information terminal 100 narrows down the loop-out information, which is information required by the user, by the narrowing-down process (S1007). Then, it responds to search for loophole information.

  As in this example, even when the user does not come up with an appropriate voice command and the information estimated by the environment information is different from the information requested by the user, the user repeatedly presses the talk button 111. Necessary information can be obtained without pressing and inputting voice.

  The example of FIG. 3C is a case where a voice recognition result cannot be obtained because the user has not come up with an appropriate voice command, as in the example of FIG. The intention estimation unit 115 estimates that information required by the user is traffic jam information. However, the example of FIG. 3C is a case where it is determined that the reliability of the estimated content is lower than a predetermined threshold.

  If it is determined that the reliability is low, the system control unit 101 generates a question content for requesting further information from the user. Here, “Would you like to know the traffic information?” Is output to confirm whether the user wants to know the traffic information.

  The user responds “Yes” because the information required when the talk button 111 is pressed is traffic jam information. In response to this, the system control unit 101 narrows down the traffic jam information by the narrowing-down process (S1007), and provides the traffic jam information obtained from the environment information.

  As in this example, even when the user does not come up with an appropriate voice command and the reliability of the information estimated from the environment information is low, the user presses the talk button 111 many times to input the voice. Without having to, you can get the information you need.

  The example in FIG. 3D is an example in the case where the user has not come up with an appropriate voice command, and the voice recognition unit 114 makes a natural utterance that is difficult to recognize.

  After the user presses the talk button 111, the system control unit 101 responds “What is it?”. Here, the user makes an utterance that is not a voice command, "How far can this traffic jam be removed?" Since this utterance is a natural utterance that is difficult for the voice recognition unit 114 to recognize the voice, a voice recognition result cannot be obtained (S1006). Therefore, the system control unit 101 acquires environment information (S1008). The intention estimation unit 115 estimates that information required by the user is traffic jam information. However, the example of FIG. 3D is a case where it is determined that the reliability of the estimated content is lower than a predetermined threshold.

  If it is determined that the reliability is low, the system control unit 101 generates a question content for requesting further information from the user. Here, “Would you like to know the traffic information?” Is output to confirm whether the user wants to know the traffic information.

  The user answers “Yes” because the question content intended when the talk button 111 is pressed is traffic jam information. In response to this, the system control unit 101 narrows down the traffic jam information by the narrowing-down process (S1007), and provides the traffic jam information obtained from the environment information.

  As in this example, even when the user utters a natural utterance that is difficult to recognize without thinking of an appropriate voice command, and the reliability of the information estimated from the environmental information is low, the user repeatedly presses the talk button 111. Necessary information can be obtained without pressing and inputting voice.

  The example of FIG. 3E is an example in the case where the user has not come up with an appropriate voice command, and thus the voice recognition unit 114 makes a natural utterance that is difficult to recognize.

  Similar to the example of FIG. 3D, the user utters a message that is not a voice command, "How long can this traffic jam be removed?" Since this utterance is a natural utterance that is difficult for the voice recognition unit 114 to recognize the voice, a voice recognition result cannot be obtained (S1006). Therefore, the system control unit 101 acquires environment information (S1008). The intention estimation unit 115 estimates that information required by the user is traffic jam information. If it is determined that the reliability is low, the information to be provided to the user and the operation to be performed cannot be estimated appropriately. Therefore, the system control unit 101 generates a question content for requesting further information from the user. Here, “Do you want to know the traffic information?” Is output to confirm whether the user wants to know the traffic information.

  However, the user recognizes that the information required when the talk button 111 is pressed differs from the provided information. Therefore, it responds “No”.

  In response to this, the system control unit 101 executes a narrowing process (S1007) excluding traffic jam information. As a result, arrival time information is estimated and provided to the user. At this time, contents for prompting the user to confirm or correct the utterance are added and output.

  The user responds “Yes” because the question content intended when the talk button 111 is pressed is arrival time information. In response to this, the system control unit 101 provides arrival time information obtained from the environment information.

  As in this example, even when the user utters a natural utterance that is difficult to recognize without thinking of an appropriate voice command, and the reliability of the information estimated from the environmental information is low, the user repeatedly presses the talk button 111. Necessary information can be obtained without pressing and inputting voice.

  FIG. 4 is an explanatory diagram showing an example of environment information.

  The environment information illustrated in FIG. 4 affects information required by the user or desired operation content. The system control unit 101 acquires this environment information and estimates information requested by the user.

  The environmental information of (1) to (6) is obtained by the system control unit 101 from the change in the position information received by the GPS receiver 102, the map information stored in the HDD 109, and the guidance information received by the traffic information receiver 104. Judge by matching.

  Whether or not there is a traffic jam (7) is determined by the system control unit 101 based on the matching result of the position information received by the GPS receiver 102 and the traffic information of the VICS received by the traffic information receiver 104 and a short history of vehicle speed. Judgment.

  Whether or not the vehicle is traveling (8) is determined based on a short time history of the vehicle speed or a side brake signal.

  Whether or not the operation is performed for a long time (9) is determined by the system control unit 101 comparing the elapsed time from the start of the engine to the present time and a preset threshold value.

  Whether or not the person is sweating (10) is determined by a signal from the sweat sensor 117 installed on the handle.

  Whether or not the currently playing CD has been played once (11) is determined by the system control unit 101 based on the repeat signal from the CD playback device of the AV device 113.

  Whether it is a meal time (12) is determined by acquiring the current time from a clock built in the system control unit 101.

  The state (13) of the road surface is determined by a signal from the road surface sensor 116.

  Whether or not the driver issues the voice command (14) is determined by the system control unit 101 by estimating the voice arrival directions of the plurality of microphones 110.

  The system control unit 101 represents the environment information as a probability distribution with respect to values that can be taken by the respective conditions.

For example, in the condition (11), possible values are binary values of true / false, and true / false judgments can be made reliably. Therefore, if the probability that the condition of (11) is true is P11 (true) and the probability that it is false is P11 (false), these values are
P11 (true) = 1, P11 (false) = 0
Or
P11 (true) = 0, P11 (false) = 1
Either.

  On the other hand, in the condition (14), the accuracy of sound source direction estimation may be reduced due to the influence of external noise. Therefore, the system control unit 101 sets the reliability of the sound source direction estimated by the system control unit 101 based on information acquired from a plurality of microphones included in the microphone 110 as a value of P14 (true).

  Also, the condition (13) is not a binary value. In such a case, the system control unit 101 sets the probability regarding each possible value as the value of the environment information. Specifically, P13 (dry), P13 (wet), P13 (water film), P13 (snow cover), and P13 (freeze) are obtained from information acquired from the road surface sensor 116.

  FIG. 5 is an explanatory diagram showing an example of information required by the user or an estimated value of the desired operation content.

  The system control unit 101 has such an estimated value in advance. Then, the degree of approximation of the estimated value is calculated from the acquired value of the environment information as shown in FIG.

  Note that setting all the information that can be provided or all the possible operations to be estimated values makes it difficult to associate with the conditions of the environment information in FIG. Therefore, it is preferable to narrow down the possible user intentions and set the estimated value.

  Next, an estimation method will be described.

  The system control unit 101 uses the environment information acquired in S1008 in FIG. 2 and the recognition result of the voice command uttered by the user to estimate information required by the user or desired operation content shown in FIG. A Bayesian network can be used for this estimation.

  FIG. 6 is an explanatory diagram of an estimation method using a Bayesian network.

  The system control unit 101 uses the environment information acquired in S1008 of FIG. 2 and the recognition result of the voice command uttered by the user as a condition. All of these conditions are set as parent nodes, and the user's intention with the type of information required by the user or the content of desired operation as a random variable is set as a child node.

  The conditional probability table for this Bayesian network is created based on a rule of thumb. Hereinafter, an embodiment of a method for creating a conditional probability table from an empirical rule will be described.

  FIG. 7 is an explanatory diagram of an example of a random variable of a Bayesian network created by an empirical rule.

  This is created in advance and stored in the information terminal 100 with regard to the user intentions (1) and (2) illustrated in FIG. 5 in consideration of the possibility that each condition causes these user intentions.

  For example, if the true / false value of the condition (7) on whether or not there is a traffic jam is “true” and the vehicle is on a toll road and is far from the toll gate, the user's intention Is more likely to need traffic information. In addition, when the position of the own vehicle is on a general road and in the vicinity of an intersection, the user's intention is likely to require exit information.

  In this way, information required by the user or desired operation content is estimated.

  As described above, for all combinations of values that can be taken by each condition, the probability that the user needs traffic information and the probability that the exit information is needed in the combination of conditions may be appropriately determined.

  On the other hand, a conditional probability table is created by the method described in the above document from actual data on what information provision requests and operation requests were made in what combinations of possible values of conditions. A Bayesian network can also be completed by an estimation method.

  As described above, the information terminal 100 according to the first embodiment of the present invention analyzes the user's voice command when the user presses the talk button 111 and provides information corresponding to the voice command. At this time, if the analysis result of the user's voice command is not obtained, the environment information is acquired, the information required by the user or the desired operation content is estimated, and the estimated content is provided. By doing so, the user can obtain necessary information or desired operation contents without pressing the talk button 110 many times and issuing voice commands many times, so that convenience for the user is improved.

  Next, a second embodiment of the present invention will be described.

  In the first embodiment described above, the user's intention is estimated after the user presses the talk button and receives the user's voice command. In contrast, in the second embodiment, the user's intention is estimated after the user presses the talk button, and information is provided to the user. Thereafter, the user's voice command is accepted.

  FIG. 8 is a flowchart of the voice interaction process of the information terminal 100 according to the second embodiment of this invention.

  First, the user presses the talk button 111 to notify the information terminal 100 of the utterance of the voice command. When the talk button 111 is pressed, a talk signal is transmitted to the system control unit 101.

  When the system control unit 101 receives the talk signal from the talk button 111 (S1001), the system control unit 101 reads the current environment information from the GPS receiver 102, the autonomous navigation sensor 103, the traffic information receiver 104, the route search unit 107, and the map information reading. Obtained from the unit 108, the road surface sensor 116, the sweat sensor 117, etc. (S1008). Then, the information required by the user or the desired operation content is estimated based on the acquired environment information (S1009).

  The intention estimation unit 115 obtains estimation contents and estimation reliability as a result of the estimation processing. Then, it is determined whether or not the reliability of the estimation is equal to or greater than a predetermined threshold (S1010).

  If the reliability is higher than the threshold, it is determined that the estimated content is equal to the user's intention. Therefore, information related to the estimated content is provided to the user (S1011). Specifically, the system control unit 101 performs display on the display 105 or audio output by the audio output device 106 and presents the information. Further, the system control unit 101 executes the estimated operation content, and notifies the user that the operation content has been executed by display on the display 105 or sound output by the sound output device 106.

  On the other hand, when the intention estimation unit 115 determines that the reliability is low in S1010, the system control unit 101 generates a question content for further information from the user (S1002). The generated question is output to the user through the display 105 and the audio output device 106 (S1003), and the user's utterance is prompted.

  Next, a voice command issued by the user (S1004) is subjected to voice recognition processing by the voice recognition unit 114, and a recognition result is obtained (S1005).

  Next, the voice recognition unit 114 determines whether or not a voice processing result is obtained as a result of the process of S1005 (S1006). If a voice recognition result is obtained, the voice recognition unit 114 uses the information included in the recognition result to narrow down information required by the user or candidates for desired operation contents (S1007).

  Therefore, when the estimation result of S1009 is the same as the user's intention, the user can obtain the necessary information or perform the desired operation simply by pressing the talk button 111.

  Similarly to the first embodiment (FIG. 2), when the estimation result is not correct, the user has an opportunity to perform correct utterance, utterance regarding the information and operation that are really necessary by the steps of S1002 and S1003. Given.

  FIG. 9 is an explanatory diagram of an example of a dialogue between the information terminal 100 and the user according to the second embodiment.

  Note that “S” represents information presentation or question by voice output from the information terminal 100, and “U” represents utterance by the user.

  In the example of FIG. 9A, when the user first presses the talk button 111, a talk signal is transmitted to the system control unit 101 (S1001). In response to this, the system control unit 101 acquires environment information (S1008). In the example of FIG. 9, the route information held by the route search unit 107, the position information from the GPS receiver 102, the position information from the autonomous navigation sensor 103, and the map information obtained from the map information reading unit 108 From the position information obtained by matching, it is found that the user is traveling on the highway and will soon reach the toll gate.

  Based on this environmental information, the intention estimation unit 115 estimates that the information required by the user is highway toll information. The reliability of the estimated content is higher than a predetermined threshold value.

  Therefore, the system control unit 101 provides highway toll information to the user as shown in FIG. After that, a question content ("Is there anything else?") That prompts the user to confirm or correct the utterance is output.

  At this time, since the user has obtained information equivalent to the question content intended when the talk button 111 is pressed, the user replies “No” and ends the dialogue with the information terminal 100.

  As in this example, when the user presses the talk button 111, the user can obtain a result equivalent to the intended question content.

  On the other hand, in the example of FIG. 9B, as in the example of FIG. 9A, after the user presses the talk button 111, the system control unit 101 provides the highway toll information to the user. After that, a question content ("Is there anything else?") That prompts the user to confirm or correct the utterance is output.

  At this time, since the user is provided with information different from the question content intended when the talk button 111 is pressed, the user responds with “Nearby convenience store” which is necessary information. In response to this, the system control unit 101 specifies the user's intention through the narrowing process (S1010), and specifies nearby convenience store information based on the acquired environment information (S1008) (S1008). This information is provided to the user (S1011).

  In the case of FIG. 9B, it is not clear whether or not the information on the high-speed fee provided first was necessary information for the user. However, if necessary, in order to acquire these two pieces of information in the conventional system, the user needs to speak a voice command at least twice (“fast charge” and “closer convenience store”). However, in the system of the present invention, the number of utterances can be reduced, and convenience is improved.

  As described above, the information terminal 100 according to the second embodiment of the present invention obtains environmental information when the user presses the talk button 111, and estimates information required by the user or desired operation content. And provide the estimated content. By doing so, the user can obtain necessary information or desired operation contents without issuing a voice command, so that convenience for the user is improved.

  In addition, even when the provided information is not what is required or what the desired operation is, the user utters more voice commands, making it easier to narrow down the user's intention and improving estimation accuracy. To do. By doing so, intended information can be obtained without the user pressing the talk button many times and issuing voice commands many times, so that convenience for the user is improved.

  The present invention can be applied to an in-vehicle information terminal such as a car navigation system and can provide a highly convenient user interface.

It is a block diagram showing the structure of the information terminal of 1st Example of this invention. It is a flowchart of the voice dialogue process of 1st Example of this invention. It is explanatory drawing of the example of the interaction of the information terminal and user of 1st Example of this invention. It is explanatory drawing which shows the example of the environmental information of 1st Example of this invention. It is explanatory drawing which shows an example of the estimated value of the information which the user of the 1st Example of this invention requires or the desired operation content. It is explanatory drawing of the estimation method using the Bayesian network of 1st Example of this invention. It is explanatory drawing of an example of the random variable of the Bayesian network of 1st Example of this invention. It is a flowchart of the voice dialogue process of 1st Example of this invention. It is explanatory drawing of the example of the interaction of the information terminal and user of 2nd Example of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Information terminal 101 System control part 102 GPS receiver 103 Autonomous navigation sensor 104 Traffic information receiver 105 Display 106 Voice output device 107 Route search part 108 Map information reading part 109 Hard disk (HDD)
DESCRIPTION OF SYMBOLS 110 Microphone 111 Talk button 112 GUI operation part 113 AV apparatus 114 Voice recognition part 115 Intention estimation part 116 Road surface sensor 117 Sweating sensor 118 Memory 119 Interface

Claims (8)

In an information terminal that recognizes a user's voice as a command and provides information based on the recognized command,
A talk signal unit that outputs a talk signal according to a user instruction;
A voice recognition unit that recognizes a voice uttered by a user as a command;
An environmental information acquisition unit that acquires environmental information about the environment around the information terminal;
An estimation unit that estimates information intended by the user from the recognized command and the acquired environment information;
A control unit for controlling processing of the information terminal;
With
The controller is
In response to reception of the talk signal, it is determined whether or not a user command recognized by the voice recognition unit has been obtained,
An information terminal characterized in that when the command cannot be obtained, the information estimated by the estimation unit is provided to a user using the environment information acquired by the environment information acquisition unit. In an information terminal that recognizes a user's voice as a command and provides information based on the recognized command,
A talk signal unit that outputs a talk signal according to a user instruction;
A voice recognition unit that recognizes a voice uttered by a user as a command;
An environmental information acquisition unit that acquires environmental information about the environment around the information terminal;
An estimation unit that estimates information intended by the user from the recognized command and the acquired environment information;
A control unit for controlling processing of the information terminal;
With
The controller is
In response to reception of the talk signal,
An information terminal characterized in that the information estimated by the estimation unit using the environment information acquired by the environment information acquisition unit is provided to a user and a command recognized by the voice recognition unit based on a user's utterance is acquired. .   The information terminal according to claim 1, wherein the control unit determines that a command cannot be acquired when a similarity between a user's utterance recognized by the voice recognition unit and a voice model held by the voice recognition unit is low.   When the control unit acquires a command recognized by the voice recognition unit, the control unit selects information corresponding to a user command using the environment information acquired by the environment information acquisition unit, and the selected information is selected by the user. The information terminal according to claim 1, wherein the information terminal is provided.   The control unit, after obtaining the command recognized by the voice recognition unit, selects information corresponding to the user's command based on the estimated content, and provides the selected information to the user. The information terminal according to claim 2.   The information terminal according to claim 1, wherein the control unit adds a question to the information estimated by the estimation unit and presents the information to the user.   The information terminal according to claim 1, wherein the control unit outputs a question asking the user for further information when the reliability of the information estimated by the estimation unit is lower than a predetermined threshold. . Provided in the vehicle,
A GPS receiver for receiving radio waves from GPS satellites;
An autonomous navigation sensor unit that includes a gyro and acquires the direction and acceleration of the vehicle by the gyro;
A map information storage unit for storing map information;
A position calculation unit for calculating the position of the vehicle;
With
The position calculation unit calculates a position on the map of the vehicle based on the radio wave received by the GPS reception unit and information acquired by the autonomous navigation sensor,
The information terminal according to claim 1, wherein the environmental information acquisition unit acquires the position of the vehicle calculated by the control unit as environmental information.

Images