JP2013037097A - Voice processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice processor capable of awaking a user in a case where the user using the voice processor for executing an interaction scenario falls into a doze.SOLUTION: A voice processor includes: a voice outputting section; a voice recognizing section; a first controller; and a first storage section that stores an interaction scenario and voice output data therein. The first controller causes the interaction scenario to progress on the basis of a voice recognition result from the voice recognizing section while causing a voice to be uttered from the voice outputting section by using the voice output data on the basis of the interaction scenario. In a scene of the interaction scenario where a response is waited, the first controller causes a voice indicating a warning to be uttered by using the voice outputting section when the voice recognition result is not transmitted to the first controller even after a predetermined time period has elapsed.

Description

  The present invention relates to a voice processing apparatus, and more particularly to a voice processing system that performs a dialogue process with a user.

  Conventionally, what is called a conversation bot or a chat bot exists as a conversation with a person. A conversation bot is a device that simulates a conversation with a person, but does not understand the content of the conversation. For this reason, conversation bots are also called artificial incompetence. For example, a conversational bot such as a talking parrot described in Patent Document 1 that speaks to a user by repeating the user's voice as it is. However, the history is old, and the conversation bot system called ELIZA, which was developed by Joseph Weisenheim in 1966, was started. In ELIZA, a therapist dressed as a therapist converts the patient's words into questions and turns them back. Thereafter, such a conversation bot system has come to be used for applications such as prevention of falling asleep when talking with an elderly person or driving a car, etc., based on the above-described treatment.

  As described above, since the conversation bot utters the same word as the word uttered by the user, there is also a drawback that the user is easily bored. In order to solve this problem, Patent Document 2 understands conversation characteristics / topics that seem to be favorable for the user and performs dialogue processing, thereby eliminating the boredom of the user such as the driver and preventing the user from falling asleep. It is described that the above can be realized.

Japanese Patent Laid-Open No. 11-9487 JP2011-125900A

  However, performing conversation characteristics / topics that seem to be favorable for the user, on the other hand, leads to a reduction in a sense of tension for the user, and in some cases the user may fall asleep. is assumed.

  SUMMARY An advantage of some aspects of the invention is to solve at least one of the problems or problems described above, and the invention can be implemented as the following application examples or embodiments.

[Application Example 1]
The voice processing device according to this application example includes a voice output unit, a voice recognition unit, a first control unit, and a first storage unit in which a dialogue scenario and voice output data are stored. The control unit causes the voice output unit to utter a voice using the voice output data based on the dialog scenario, and advances the dialog scenario based on the voice recognition result from the voice recognition unit, When the voice recognition result is not transmitted to the first control unit even after a predetermined time has elapsed in a response waiting scene, the first control unit uses the voice output unit to output a voice indicating a warning. It is characterized by uttering.

  According to this configuration, the voice processing device includes a voice output unit, a voice recognition unit, a first control unit, and a first storage unit in which a dialogue scenario and voice output data are stored. When the voice recognition result is not transmitted to the first control unit even after a predetermined time elapses in a response-waiting scene, the first control unit uses the voice output unit to indicate a warning. , It is possible to call attention to a person who has fallen asleep or is about to fall asleep while using the voice processing device.

[Application Example 2]
In the audio processing device according to the application example, the first control unit includes a measurement counter for measuring the predetermined time, and a count value for measuring the predetermined time is a value waiting for the response. It is preferably set for each scene.

  According to this configuration, the first control unit has a measurement counter for measuring a predetermined time, and sets a count value for measuring the predetermined time for each response-waiting scene. A waiting time can be set according to the waiting scene. There are various scenes waiting for a response in the dialogue scenario, such as a scene where the user can respond immediately and a scene where the user needs to think or confirm. For this reason, for example, when the waiting time is set simply by the average time of the user's response, if a dialogue that requires thought appears after a dialogue that can be answered immediately, the waiting time may elapse during thinking. There is sex. By setting the waiting time in the response waiting scene in consideration of the thinking time according to the scene, it is possible to set a more natural and preferable conversation environment for the user.

[Application Example 3]
In the speech processing apparatus according to the application example described above, it is preferable that a maximum value that can be measured by the measurement counter is set as a count value for the predetermined time in the first response waiting scene in the conversation scenario.

  According to this configuration, a situation occurs in which the predetermined time elapses in the first response waiting scene by setting the predetermined time as the maximum time that can be measured by the measurement counter in the first response waiting scene in the dialogue scenario. It is possible to reduce the possibility of doing so. The first waiting for a response when the conversation scenario starts is less likely to be asleep or bored than the other scenes. Can be expected to respond at the right time. Therefore, even if the predetermined time setting is the maximum value that can be measured by the measurement counter, it is unlikely that the measurement counter will count up, it can be determined that there is no effect on the actual progress of the dialogue scenario, and an inadvertent warning is issued. Can be prevented.

[Application Example 4]
In the speech processing apparatus according to the application example, the predetermined time may be changed based on a time required for the first control unit to recognize the speech recognition result in the response waiting scene. preferable.

  According to this configuration, the change of the predetermined time is changed based on the time required until the voice recognition result is transmitted after the transition to the response waiting scene. Can be the appropriate length of time. As described above, the predetermined time varies depending on the response waiting scene. There are multiple waiting times for a response in a dialogue scenario. For example, the waiting for a response is a waiting for a response that can be answered immediately by the user or a waiting for a response that needs to be answered by the user. Multiple types of scenes are possible. By changing the count value of the predetermined time in each response waiting scene based on the time required for the first control unit to recognize the speech recognition result for each type, the response of the device is further improved. It can be suitable for the user.

[Application Example 5]
In the speech processing apparatus according to the application example described above, it is preferable that the first control unit stores a history of the speech recognition result in the first storage unit in association with the conversation scenario.

  According to this configuration, the first control unit stores the voice recognition result in the first storage unit in association with the dialogue scenario, so that the voice recognition result can be used as a history of the dialogue scenario. The history can be used for later treatment when the user is a patient.

[Application Example 6]
In the speech processing apparatus according to the application example described above, it is preferable that there are a plurality of conversation scenarios having different contents, and a dialog person can select which conversation scenario to use.

  According to this configuration, by preparing a plurality of dialogue scenarios and allowing the user to select which dialogue scenario to use, it is possible to prevent the user from repeatedly using the same dialogue scenario. The plurality of dialogue scenarios may be stored in the first storage unit from the beginning, or may be newly stored in the first storage unit using an external storage device or a network. In any case, having a plurality of dialogue scenarios can prevent the user from getting bored with the use of the device.

1 is a schematic block diagram of an audio processing device. A part of flowchart of the process in 1st Embodiment. A part of flowchart of the process in 1st Embodiment. An example of a chart diagram of an audio scenario. An example of a chart diagram of an audio scenario. 1 is a schematic block diagram of an embodiment of a voice processing device.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the drawings used in the description of the present embodiment, for the convenience of description, description of parts unnecessary for description is omitted or simplified. In the following description, binary data is expressed by adding b to the end of numerical values, and hexadecimal data is expressed by adding h to the end of numbers.

(First embodiment)
FIG. 1 shows a schematic block diagram of a speech processing apparatus 10 in the present embodiment. The voice processing device 10 is a device having a dialog bot function, and is connected to a microphone 21 and a speaker 22. The user listens to the sound output from the speaker 22 and talks to the microphone 21 to interact with the sound processing apparatus 10. In the present embodiment, it is assumed that the microphone 21 and the speaker 22 are appropriately interfaced with the sound processing apparatus 10.

  The voice processing device 10 includes a voice recognition unit 11, a first control unit 12, a first storage unit 13, and a voice output unit 14. Further, the first control unit 12 has a measurement counter 31. The first storage unit 13 stores a dialogue scenario, voice data for synthesizing voice output from the speaker 22, data necessary for voice recognition (voice feature data), and the like. It is assumed that a plurality of dialogue scenarios are prepared.

  The voice recognition unit 11 has a function of processing the voice taken in by the voice processing device 10 and converting it into a predetermined symbol string. Although not shown, the speech recognition unit 11 includes an AD converter that converts an analog signal from the microphone 21 into a digital signal. The speech recognition unit 11 extracts a meaningful phrase from a predetermined symbol string of the digitally converted speech and propagates it to the first control unit 12. Voice feature data stored in the first storage unit 13 is used to extract the meaningful phrase.

  The voice output unit 14 has a function of synthesizing voice data defined in the dialogue scenario in accordance with an instruction from the first control unit 12 and outputting voice through the speaker 22.

  The first control unit 12 is a part that controls the voice recognition unit 11 and the voice output unit 14 based on the dialogue scenario stored in the first storage unit 13. When the voice processing device 10 is turned on by an operating unit (not shown) and a start of dialogue is instructed by a user operation, the voice processing device 10 refers to the dialogue scenario in the first storage unit 13 and recognizes the voice recognition unit 11. The dialogue scenario is advanced by causing the voice output unit 14 to perform necessary processing.

  FIG. 2 shows a part of the processing in the speech processing apparatus 10 as a flowchart 100. When the user turns on the power of the speech processing apparatus 10, a power-on reset as a hardware operation is executed. Thereafter, necessary initial settings are made in the speech processing apparatus 10 (S001). The initial setting is to set an operation mode or the like necessary for the voice processing apparatus 10 to perform an appropriate operation, whereby the first control unit 12, the voice recognition unit 11, and the voice output unit 14 have their respective functions. You will be in the right condition to fulfill.

  Next, the user interface (UI) is activated, and the user can operate the speech processing apparatus 10 (S002). As a result, the voice processing apparatus 10 waits for an instruction from the user (S003). In the present embodiment, for convenience of explanation, it is assumed that the user's operation is limited to ending the processing of the voice processing device 10 or selecting a dialogue scenario. Although not particularly mentioned, the part for executing the initial setting or the UI activation may be performed by a component of the voice processing apparatus 10 (not shown in FIG. 1) or may be performed by the first control unit 12.

  When an operation is performed by the user, it is first determined whether or not it is an end instruction (S004). In the case of an end instruction, a predetermined end process is performed and the processing of the voice processing device 10 ends. If it is not the end process, the dialog scenario is selected by the user, and the scenario selection process is performed (S005). Next, the first control unit 12 reads the data of the corresponding scene (in this case, the start scene) of the corresponding dialog scenario, and thereafter performs processing necessary for the progress of the dialog scenario (S006, S007, S009, etc.). If the corresponding scene is a scene where audio is output, an audio output process is executed (S008), and it is determined whether the scenario has ended (S015). If not, the next scene is executed. The process returns to the scenario progress process (S006). If the corresponding scene is a scene where voice input is performed, the process proceeds to voice recognition processing (S010 to S014). If not, it is determined whether the scenario is finished (S015). It returns to processing (S006). If it is determined that the dialogue scenario has been completed (S015), the process returns to a state of waiting for a user instruction (S003).

  The voice recognition process (S010 to S014) in the first control unit 12 is executed as follows. First, the first control unit 12 turns on the operation of the voice recognition unit 11, sets a predetermined value in the measurement counter 31 for monitoring the response waiting time, and starts the measurement counter 31 (S010). Next, interruption of the voice recognition unit 11 and the measurement counter 31 is permitted (S011). Thereafter, the processing flag that is turned on in the interrupt processing of the voice recognition unit 11 is monitored (S012), and the count-up flag that is turned on in the interrupt processing of the measurement counter 31 is monitored (S013).

  When the processing flag is on, there is a response from the user, indicating that the speech recognition processing has been performed normally. In this case, it is determined whether or not the scenario is completed (S015). If not completed, the process returns to the scenario progress process (S006). When the count-up flag is on, there is no response from the user within the time set as the response wait time. In this case, a warning process (S014) for prompting the user to respond is performed. Thereafter, since the interrupt processing of the measurement counter 31 is completed in a state where interrupt is not permitted, the interrupt is permitted (S011), and the above processing (S012, S013) for monitoring the end of the interrupt processing is repeated.

  The state of monitoring the interrupt process (S012, S013) in a normal manner upon the progress of the dialogue scenario is exited when it is determined that the processing flag is on (S012). This indicates that there was a response from the user and that the content of the response could be recognized before a predetermined time. In this case, the process proceeds to S015 of the flowchart 100 in order to advance the dialogue scenario.

  The reason why the interruption process monitoring (S012, S013) exits in an abnormal manner in the progress of the dialogue scenario is that the voice recognition result by the response from the user passes a predetermined time measured by the measurement counter 31. This is when you cannot get it. This is the case when the count-up flag is turned on by the interrupt processing of the measurement counter 31. In this case, it is determined in S013 of the flowchart 100 that the flag is turned on, and the process proceeds to a warning process (S014). The warning process (S014) will be described later. Since the warning process (S014) is performed in the interrupt mask state, after the warning process (S014), the process proceeds to interrupt permission (S011), shifts to a state waiting for the user's reply, and monitors the end of the interrupt process. The above processing (S012, S013) is repeated. If the voice recognition result of the response from the user cannot be obtained, the process for monitoring the end of the interrupt process and the warning process are repeated.

  The interruption from the voice recognition unit 11 occurs when a predetermined phrase to be recognized by the voice recognition unit 11 is recognized. When this interrupt occurs, the first control unit 12 determines that there is a response from the user and that the response has been recognized normally.

  A schematic flowchart of the interrupt process of the voice recognition unit 11 is shown in FIG. When entering the interrupt process, first, the interrupt is masked so as not to be a multiple interrupt (S101). Next, the first control unit 12 confirms the speech recognition result of the speech recognition unit 11 (S102), and stores the result in a predetermined area of the first storage unit 13 (S103). The recognition result is stored in association with the scene number (explained later) of the dialogue scenario, and can be used as a history by reconstructing the dialogue scenario and the recognition result after completion of the dialogue scenario. Become. Thereafter, the processing flag indicating the end of the speech recognition processing is turned on (S104), and the interrupt processing of the speech recognition unit 11 is ended. As described above, when the processing flag is determined to be ON in S012 of the flowchart 100, the process goes through the interrupt monitoring loop constituted by the processing of S012 and the processing of S013 and proceeds to the determination of the end of the conversation scenario (S015). become.

  Next, interrupt processing of the measurement counter 31 will be described. The interruption of the measurement counter 31 occurs when the count of the measurement counter 31 is counted up. The occurrence of this interrupt means that the user's response has not passed a predetermined time in a response waiting scene, and indicates a state in which the user is assumed to be in a dozing state. A schematic flowchart of the interrupt processing of the measurement counter 31 is shown in FIG. When entering this interrupt process, first, interrupts are masked so as not to be multiple interrupts (S201). Next, the count-up flag is turned on (S202), and the interrupt process is terminated. As described above, whether or not the count-up flag is on is monitored in S013 of the flowchart 100. If the count-up flag is on, the process proceeds to a warning process (S014).

  A schematic flowchart of the warning process is shown in FIG. The first control unit 12 shifts the scene of the dialogue scenario to a warning scene (S401). The warning scene will be described later. Next, an instruction to reproduce the warning sound set in the warning scene is given to the voice output unit 14 (S402), the count value is reset in the measurement counter 31 (S403), and the warning process is terminated.

  In addition, when the first control unit 12 instructs the start of the operation (S010), the voice recognition unit 11 starts analyzing the sound input from the microphone 21. A schematic flowchart of the processing of the speech recognition unit 11 is shown in FIG. The sound from the microphone 21 is converted into digital data by the AD converter of the voice recognition unit 11, and the digital data is processed in the voice recognition unit 11 as a predetermined symbol string. The predetermined symbol string is acquired as recognition data (S301), and the recognition data is analyzed by comparing the feature in the recognition data with the voice feature data stored in the first storage unit 13 (S302). Next, it is determined whether or not there is an authentication phrase corresponding to the scene of the dialogue scenario in the recognition data (S303). If there is no authentication phrase, acquisition of authentication data is continued (S301), and the authentication phrase can be extracted. Is held in a predetermined register (S304), an interrupt signal is generated (S305), and the speech recognition process is terminated. The first control unit 12 acquires, as a recognition result, the authentication phrase held in S304 of the speech recognition process (FIG. 3- (c)) in the interrupt process (FIG. 3- (a)) of the speech recognition unit 11. (S102, S103). The authentication phrase will be described later.

  Next, a dialogue scenario will be described. The conversation scenario may be considered as scene information describing the execution contents of the scene linked in the order of conversation progress. As an example, FIGS. 4 and 5 show one form of dialogue scenario. In this embodiment, three types of scene information are defined. The first piece of scene information is scene information (hereinafter referred to as first scene information) of a scene (hereinafter referred to as a first scene) in which the voice processing apparatus 10 talks to the user. The first scene information is indicated by DS001, DS003, and DS101 in FIGS. The second piece of scene information is scene information (hereinafter referred to as second scene information) of a scene where the user responds, that is, a scene where the speech processing apparatus 10 performs voice recognition (hereinafter referred to as the second scene). is there. The second scene information is indicated by DS002 in FIGS. The third scene information is the scene information (hereinafter referred to as the third scene) of the scene after the predetermined time has elapsed in the scene in which the voice processing device 10 is waiting for a response from the user (hereinafter referred to as the third scene). Called information). The third scene information is indicated by DS002W in FIG. In addition, the warning scene mentioned above is a 3rd scene. Each scene information will be described below. The specific data format of each scene information is not specifically mentioned.

  The first scene information includes information on a scene number (scene number), a voice output flag, a voice recognition flag, and a voice phrase. The scene numbers in the first scene information and the second scene information are assigned by a common rule, and are represented by a number that is continuous with an alphabet representing the type of dialogue scenario. In this embodiment, for convenience, it is composed of one alphabetic character and four digits. Basically, the progress of the scenes in the dialogue scenario is performed in the order of scenes where the number increases by 1 in the same alphabetic character. That is, after the scene No. A0100 is executed, the scene No. A0101 is basically executed.

  The audio output flag is a flag for instructing audio output. In the present embodiment, it is defined as a 2-bit code “01b”. The input speech is analyzed by the speech recognition unit 11. The voice phrase defines the voice to be output.

  The second scene information includes information of scene No., voice output flag, voice recognition flag, counter setting value, and authentication phrase / scene No. The definitions of the scene number, the voice output flag, and the voice recognition flag are the same as the first scene information. The counter setting value is a count value for measuring the response waiting time, and is set in the measurement counter 31. When the count counter 31 is instructed to start counting (S010), the measurement counter 31 is counted with a predetermined clock. The count may be either increment or decrement, and the counter set value may be determined in consideration of the frequency of the predetermined clock and the usage form of the measurement counter 31. Further, the count-up of the measurement counter 31 may be detected by carry or borrow, and a count-up may be defined by providing a comparison register or the like and matching with the value of the comparison register. As described above, the count-up flag is turned on in the interrupt process by counting up the measurement counter 31, and the warning process (S014) is executed when the flag is determined to be turned on in S013 of the flowchart 100.

  The information of the authentication phrase / scene number indicates the authentication phrase that is determined to be valid in this scene and the scene number of the scene that is executed next when the authentication phrase is recognized. For example, DS002 in FIG. 4 defines that when “Yes” is recognized as the voice recognition result of the voice recognition unit 11, the scene number of the scene to be executed next is A0102. Similarly, DS002 defines that the scene number of the scene to be executed next is B0100 when the voice recognition result of the voice recognition unit 11 is “No”.

  The third scene information includes information of scene No., voice output flag, voice recognition flag, counter setting value, voice phrase, and authentication phrase / scene No. The third scene information includes both information in the first scene information and information in the second scene information. The definition of information other than the scene No. is the same as the contents described in the first scene information and the second scene information. The scene number of the third scene information is defined as a code in which W is added to the end of the scene number of the scene waiting for a response. As described above, the third scene information is a scene in the warning process (S104). After issuing a warning to the user, the user's response must be recognized in the same manner as in the second scene immediately before. Therefore, since it is a scene where both the voice is emitted and the response is recognized, both the information in the first scene information and the information in the second scene information are included. The information of the authentication phrase / scene number is the same as that in the second scene immediately before.

  The process flow of the dialogue scenario after DS001 in FIG. 4 is as follows.

  First, in S006 of FIG. 2, the information of DS001 of FIG. 4 is read by the first control unit 12 and the scene information is interpreted. Since the first control unit 12 has a voice output flag set (S007 in FIG. 2), the first control unit 12 outputs the voice data “Let's talk about the past” in the voice phrase to the voice output unit 14. (Included in S008 of FIG. 2). The audio output unit 14 outputs the audio data in accordance with an instruction from the first control unit 12 (included in S008 of FIG. 2).

  Since there is scene information DS002 next to DS001 in the dialogue scenario (S015 in FIG. 2), the first control unit 12 reads the information in DS002 and interprets the scene information (S006 in FIG. 2). Since DS002 does not have a voice output flag (S007 in FIG. 2) and has a voice recognition flag (S009 in FIG. 2), the first control unit 12 starts the operation of the voice recognition unit 11 and The value of the counter set value is set in the measurement counter 31 to start the measurement counter (S010 in FIG. 2). Subsequently, the first control unit 12 permits interruption from the voice recognition unit 11 and the measurement counter 31 (S011 in FIG. 2), and enters a state of waiting for confirmation of various flags set by the interruption process (S012 in FIG. 2). , S013). This state is also included in the state in which the second scene (scene number A0101) is being executed.

  In the second scene (scene No. A0101), it is assumed that a response is received from the user that “the old story is not good” before the measurement counter 31 counts up. The voice recognition unit 11 analyzes the content of the recognition data of the response from the user (S301 in FIG. 3- (c)) (S302 in FIG. 3- (c)), and corresponds to the authentication phrase defined in DS002. If there is a corresponding one after confirming whether there is any (S303 in FIG. 3- (c)), the confirmed authentication phrase is extracted from the recognition data (see FIG. 3- (c)). S304), the extracted phrase is held in a predetermined location accessible by the first control unit 12. In the case of this reply, the phrase “No” is extracted. Thereafter, the voice recognition unit 11 generates an interrupt and ends the process.

  When the interruption from the voice recognition unit 11 occurs, the voice recognition interruption process (FIG. 3- (a)) is executed, and the processing flag is turned on (S104 in FIG. 3- (a)). When it is confirmed that the processing flag is turned on (S012 in FIG. 2), it is confirmed whether or not the executed scene is an end scene (S015 in FIG. 2). The process proceeds to S006). In the case of this dialogue scenario, since the phrase recognized in the scene No. A0101 is “No”, the next scene to be executed is the scene No. as described in DS002 of FIG. This is a scene of B0100. The first control unit 12 executes the first scene indicated by DS101 in FIG.

  Next, in the second scene (scene No. A0101), a case where a count-up interruption of the measurement counter 31 occurs before the voice recognition unit 11 extracts an authentication phrase will be described. In this case, the count-up flag is turned on in the measurement counter interruption process (FIG. 3- (b)) (S202 in FIG. 3- (b)). When the count-up flag is turned on (S013 in FIG. 2), a warning process (S014 in FIG. 2) is executed. Since this is a warning process in the second scene (scene No. A0101), the scene No. of the third scene to be executed is A0101W (DS002W in FIG. 5).

  Since the voice output flag of DS002W is “01b”, the first control unit 12 instructs the voice output unit 14 to output the voice phrase “Waking up” defined in DS002W (FIG. 3). (D) S402). The first control unit 12 sets the counter setting value defined in DS002W in the measurement counter 31 (S403 in FIG. 3D), and ends the warning process. Since the scene with the scene number A0101W is a scene that is understood as an extension of the scene with the scene number A0101, the speech recognition flag is set to “10b”.

  If there is no response after a predetermined time, the warning process is executed as described above, the voice phrase defined in the third scene information is output, and the user can be alerted. . Thereby, when the user is in a dozing state or is in a state of going to fall asleep, an effect of awakening the user can be expected.

  Also, by setting the maximum time that can be measured by the measurement counter in the first response waiting scene in the dialogue scenario, there is a possibility that the predetermined time will elapse in the first response waiting scene. Reduction can be achieved. The first waiting for a response when the conversation scenario starts is less likely to be asleep or bored than the other scenes. Can be expected to respond at the right time. Therefore, even if the predetermined time setting is the maximum value that can be measured by the measurement counter, it is unlikely that the measurement counter will count up, it can be determined that there is no effect on the actual progress of the dialogue scenario, and an inadvertent warning is issued. Can be prevented.

  Note that the time required for replying to the same question of the voice processing device 10 varies depending on the person. Therefore, by changing the setting value for the measurement counter 31 as the progress of the dialogue scenario proceeds, the time until the time is up can be changed to a value suitable for the user. As described above, the response waiting time is set for each second scene. The average of the counter setting value set in the second scene of the predetermined type and the time required for speech recognition in the second scene of the predetermined type among the plurality of types is determined next time. By setting the counter setting value of the second scene, it is possible to make the response waiting time measurement in the second scene of the predetermined type suitable for the user.

  Although the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment. For example, various methods are conceivable for defining scene information of a dialogue scenario. Further, the voice processing device 10 may be configured in the form of a microcomputer (voice processing device 50) as shown in FIG. The audio processing device 50 includes a CPU unit, a reset control unit (Reset), a clock generation unit (Clock), a system bus 60, a work memory unit, a timer unit 51, a peripheral unit 52, a storage unit 53, an analog-digital converter (A / D conversion) and a digital-analog converter (D / A conversion). The present invention is not limited to the application examples and embodiments described above, and can be widely applied without departing from the spirit of the present invention.

  DESCRIPTION OF SYMBOLS 10 ... Voice processing apparatus, 11 ... Voice recognition part, 12 ... 1st control part, 13 ... 1st memory | storage part, 14 ... Sound output part, 21 ... Microphone, 22 ... Speaker, 31 ... Measurement counter, 50 ... Voice processing apparatus , 51 ... timer part, 52 ... peripheral part, 53 ... storage part, 60 ... system bus, 100 ... flowchart.

Claims (6)

An audio output unit;
A voice recognition unit;
A first control unit;
A first storage unit storing a dialogue scenario and voice output data,
The first control unit utters speech from the speech output unit using the speech output data based on the conversation scenario, and advances the conversation scenario based on a speech recognition result from the speech recognition unit,
When the voice recognition result is not transmitted to the first control unit even after a predetermined time has elapsed in the response waiting scene in the dialogue scenario, the first control unit uses the voice output unit to warn A voice processing apparatus that utters a voice indicating The first control unit has a measurement counter for measuring the predetermined time,
The audio processing apparatus according to claim 1, wherein a count value for measuring the predetermined time is set for each scene waiting for a response.   3. The voice processing apparatus according to claim 1, wherein, in the first waiting for response in the dialogue scenario, a maximum value that can be measured by the measurement counter is set as the count value for the predetermined time. .   4. The predetermined time is changed based on a time required for the first control unit to recognize the voice recognition result in the response waiting scene. The voice processing device according to claim 1.   5. The voice according to claim 1, wherein the first control unit stores a history of the voice recognition result in the first storage unit in association with the dialogue scenario. 6. Processing equipment.   6. The voice processing according to claim 1, wherein there are a plurality of conversation scenarios having different contents, and a conversation person can select which conversation scenario to use. apparatus.

Images