JP2003108191A - Voice interacting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice interacting device which does not lower operation efficiency. SOLUTION: When voice interaction input is done during operation, a cognitive information processing load increases and the operation efficiency decreases. For the purpose, the voice interacting device having a voice recognition part 12 is provided with an operation state detection part 20, an operation load estimation part 25, and an interaction control part 13. The voice recognition part 12 changes a language dictionary for recognition and a sound dictionary for recognition according to the operation load. Consequently, the recognition rate is improved. Further, the interaction control part 13 makes a question answer sentence and a confirmation answer sentence single when it is estimated that the operation load is heavy, and allows a plurality of the question answer sentences or a question answer sentence containing implicit confirmation when the operation load is light. Consequently, the cognitive information processing load by interaction input is always suitably controlled and the operation efficiency does not decrease. Namely, the operation and interaction input are both enabled by changing the interaction systems.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice interactive apparatus which enables interactive input even in various worker states under various environments. In particular, the present invention relates to a voice dialogue device that makes work and dialogue input compatible even under various conditions. INDUSTRIAL APPLICABILITY The present invention can be applied to a voice interaction device that grasps a vehicle type or a running state of a vehicle and a driver's state and accurately inputs a dialogue.

[0002]

2. Description of the Related Art Conventionally, there is a voice dialogue device which enables dialogue input under various environments. For example, Japanese Patent Laid-Open No. 10-2
No. 0884, Japanese Patent Laid-Open No. 10-104
No. 009 discloses a navigation device. The former is, as shown in FIG. 11, a voice input unit 101 and a voice recognition unit 10.
2, an example in which a dialogue management unit 106, a dialogue storage unit 105, and a proficiency level detection unit 110 are added to a conventional voice dialogue device including a dictionary storage unit 103, a dictionary selection unit 104, a guidance selection unit 107, and a voice output unit 109. is there. Briefly, when the driver inputs a voice from the voice input unit 101,
The voice recognition unit 102 performs voice recognition. The dictionary used for voice recognition is the recognition dictionary selected by the dictionary selection unit 104 from the dictionary storage unit 103. Then, the dialogue management unit 10
6 manages the flow of the dialogue according to the stored contents of the dialogue storage unit 105 and the recognition result of the voice recognition unit 102. The skill level detection unit 110 determines the skill level of the user based on the speed of interaction. Then, the guidance selection unit 107 determines the voice guidance to be output from the guidance storage unit 106 according to the result (proficiency level) of the skill level detection unit 110 and the stored content in the dialogue storage unit 105. Finally, it is output from the audio output unit 109. As described above, JP-A-10-20884
The speech dialogue device disclosed in the issue is characterized in that the contents of the guidance are changed according to the degree of proficiency so that the dialogue can be carried out efficiently.

The latter conventional example is a vehicle-mounted navigation device, which is a device that applies various restrictions to the provision of voice information depending on the running state of the vehicle. For example, it is characterized in that the provision of voice information is stopped when the vehicle makes a sharp turn or is suddenly braked. Further, it is characterized in that only voice information of the registered characteristic media is provided when the vehicle suddenly turns or is suddenly braked. Another feature is that the order of the audio information provided is changed. It is also characterized in that the total amount of voice information is limited and the time required for it is regulated. Further, when the provision of the voice information is stopped, the voice information is provided as the character information.

[0004]

SUMMARY OF THE INVENTION As described above, Japanese Patent Laid-Open No.
The voice interaction device disclosed in No. 0-20884 is an example in which the voice response is changed according to the skill level of the worker. That is, it does not take into consideration the work state such as environmental noise around the worker or the state of the work device. For example, when the voice interaction device is applied to a vehicle, a vehicle engine sound, wind noise, road noise, etc. may be mixed in the voice input. In such a case, voice recognition becomes difficult. That is, voice dialogue may be difficult. In addition, an operator (driver) may be operating another device during voice interaction. For example, the vehicle may be accelerating or braking.
In such a case, when the voice dialogue is performed, the attention is lowered due to an increase in the cognitive information processing load due to the voice dialogue, and the vehicle operation may be delayed.

The navigation device disclosed in Japanese Unexamined Patent Publication No. 10-10409 is an example in which the running state of the vehicle is used as a parameter, and the voice conversation is variously regulated by the parameter. For example, it is an example in which the voice conversation is completely stopped depending on the running state. That is, the vehicle operation and the voice dialogue are not completely compatible with each other. Further, the above navigation system is an example in which only the running state of the vehicle is used as a parameter, and the driver (worker) state is not used as a parameter. For example, it is known that when the driver concentrates (tenses) on driving, the utterance characteristics of the driver change and the voice recognition rate decreases. That is, in order to carry out an accurate voice dialogue, it is necessary to grasp the driver's state (the driver's heart rate, eye movement, etc.) and change the voice recognition condition. That is, the navigation system that does not use the driver's state as a parameter does not always guarantee an accurate voice dialogue.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to appropriately control a voice dialogue system, a recognition dictionary, etc. based on a work load of an operator. It is possible to enable interactive input in various environments and in various worker states without lowering work efficiency. That is, it is to make work and dialogue input compatible. Another object of the present invention is to apply the voice dialogue system of the present invention to a voice dialogue system in a passenger compartment of an automobile so as to realize smooth voice dialogue while maintaining driving safety. The purpose of these inventions should not be understood as that each invention should achieve all the objects, and the corresponding individual inventions are:
Each should be understood as an objective to be achieved.

[0007]

A voice dialog device according to claim 1 is a voice dialog device which enables a dialogue input during a work by using a voice recognition technique. Using a work load estimation unit that estimates the work load of the worker based on the detection result of the state detection unit, one or more recognition language dictionaries, and one or more recognition acoustic dictionaries, Estimated by the speech recognition unit that recognizes the input, the recognition result of the speech recognition unit, the database search result corresponding to the recognition result, the conversation progress information composed of the database search result and the certainty, and the work load estimation unit. A dialogue control unit that selects a confirmation item and / or a question item and a voice feature that does not increase the work load of the worker based on the work load level, and a confirmation guidance to the worker based on the result of the dialogue control unit. Characterized in that a guidance generator for generating a beauty / or questions guidance.

The speech dialogue system according to a second aspect of the invention is the speech dialogue system according to the first aspect, wherein the speech recognition section is based on the estimation result of the work load estimation section, and a plurality of recognition language dictionaries are provided. And an optimal recognition language dictionary and a recognition acoustic dictionary are selected from a plurality of recognition acoustic dictionaries for recognition. Further, the voice dialogue apparatus according to claim 3 is the voice dialogue apparatus according to claim 1, wherein the voice recognition unit is based on the detection result of the work state detection unit and / or the dialogue progress degree information of the dialogue control unit. , A plurality of recognition language dictionaries and a plurality of recognition sound dictionaries are selected to recognize an optimal recognition language dictionary and recognition sound dictionary.

Further, the voice dialogue apparatus according to a fourth aspect is the voice dialogue apparatus according to any one of the first to third aspects, wherein the work state detecting section detects the worker state. It is characterized by comprising a state detection unit, and / or a device state detection unit that detects the state of the work device, and / or a noise detection unit that detects ambient noise.

Further, a voice dialogue system according to a fifth aspect is the voice dialogue system according to the fourth aspect, wherein the worker state detecting section detects the worker state from the physiological index of the worker. To do. A sixth aspect of the present invention is the voice interaction device according to the fifth aspect, wherein the physiological index of the worker is a heart rate, a sweat rate, and a line-of-sight related value.
Further, the voice interaction device according to claim 7 is the voice interaction device according to any one of claims 4 to 6, wherein the device state detection unit is based on the physical quantity of the work device and the frequency of use of the device. It is characterized in that the device state is detected.

Further, the voice dialogue apparatus according to claim 8 is the voice dialogue apparatus according to any one of claims 1 to 7, wherein the dialogue progress information is confirmation items and confirmation of the confirmation items. One of the unknown state, the confirmation required state, and the confirmed state indicating the degree 1
It is characterized by being composed of two. Further, the voice dialogue apparatus according to claim 9 is the voice dialogue apparatus according to any one of claims 1 to 8, and the dialogue control unit requires confirmation when the workload is small or medium. It is characterized by using a confirmation item that is in a state to prompt confirmation of the item and setting a new question item.

Further, a voice dialogue system according to a tenth aspect is the voice dialogue system according to any one of the first to eighth aspects, wherein the dialogue control section has a large work load, It is characterized in that the question items are set in a single question-and-answer format. Further, the voice dialogue apparatus according to claim 11 is the voice dialogue apparatus according to any one of claims 1 to 10, wherein the dialogue control unit has a predetermined time and / or a workload after a predetermined time. If it is estimated to be large, it is characterized by adjusting the interaction method, the interaction time interval and / or the utterance length, and the utterance speed so that the interaction load does not increase for the predetermined time and / or after the predetermined time.

Further, the voice interactive apparatus according to claim 12 is the voice interactive apparatus according to any one of claims 1 to 11, wherein the work state detecting section and the work load estimating section are voice recognizing sections. Also, the process is independent of the dialogue control unit and the guidance generation unit.

A voice dialogue system according to a thirteenth aspect is the voice dialogue system according to any one of the first to twelfth aspects, wherein the working device is a vehicle and the worker drives the vehicle. It is characterized by being a person. The voice interaction device according to claim 14 is the voice interaction device according to claim 13, wherein the physical quantity of the work device is a vehicle-to-vehicle distance between a front vehicle and / or a rear vehicle, a vehicle position, a speed, an acceleration, and a turn. It is characterized by including at least one of speed, turning acceleration, and accelerator / brake pedal depression amount. Further, a voice dialogue device according to claim 15 is the voice dialogue device according to claim 13 or 14, wherein the voice dialogue device is provided in an information providing device mounted on a vehicle. .

[0015]

ACTION AND EFFECT This section mainly describes the action and effect of the invention described in each claim. For facilitating the understanding of the invention, the invention is embodied and described as an example, but the structure of the claims is not limited thereto. The parts that have been concretely illustrated and described are also descriptions of the embodiments of the present invention. According to the voice interaction device of the first aspect, in the voice interaction input during work, the work state detection unit first detects the work state. The work state is a margin, a degree of complexity, a driving state of the work device, or the like related to the work load of the worker. Then, the work load estimating unit estimates the work load of the worker based on the detection result of the work state detecting unit. The workload means a cognitive information processing load, and the small workload means that there is a large margin to accept a new cognitive information processing load. On the contrary, a large work load means that the margin for accepting a new cognitive information processing load is small. This work load is detected and updated, for example, every predetermined time. The voice recognition unit has a single or plural recognition language dictionary, and a single or plural recognition acoustic dictionary,
The utterance of the worker is recognized by referring to the recognition language dictionary and the recognition acoustic dictionary.

Then, the dialogue control unit provides the recognition result of the voice recognition unit, the database search result (including the presence or absence of the database) corresponding to the recognition result, and the dialogue progress degree information based on the search result and the degree of determination. Create. The dialogue progress information includes dialogue items based on the recognition result or the search result and the degree of confirmation thereof. For example, if the working device is a vehicle and the dialog input target is a navigation device, the dialog items are the name of the destination, the business type, the address, the detailed address, and the like. The degree of confirmation is the degree of confirmation, and is displayed as 1, 2, or 3, for example.
Then, the determined number (the number of items with the degree of determination of 3) is the degree of progress. Further, the dialogue control unit selects the next confirmation item and / or question item and voice feature so as not to increase the information processing load of the worker, in consideration of the work load level estimated by the work load estimation unit. Output to the guidance generation unit of the column. For example, when the work load is large, the following confirmation item or question item is output as a single item. In the next stage, for example, is the place'Nagoya City '? Becomes a single sentence question. The worker (driver) responds with, for example, “high” or “yes”, so that the cognitive information processing load becomes small. Therefore, work efficiency (driving operation) is not reduced. On the contrary, when the work load is small, the dialogue control unit outputs, for example, a plurality of confirmation items and question items. At this time, in the next stage, for example, 'Nagoya City'
It becomes the question sentence "What kind of restaurant is it?"
In this case, 'Nagoya City' is the last recognition result and database search result and is an implicit confirmation item. The question item is "What kind of restaurant is it?"
In this case, since the work load is small, the work efficiency is not lowered even by recognizing a plurality of items.

Finally, the guidance generation section outputs the above-mentioned confirmation guidance and / or question guidance by voice synthesis of the selected voice feature based on the output of the dialogue control section. The operations of the voice recognition unit and the dialogue control unit are repeated until all the items of the dialogue progress information are determined. In this way, the dialogue control unit changes the dialogue item depending on the work state, so that the work efficiency does not decrease. Moreover, the work operation is not stopped. That is,
Work and voice dialogue input can be compatible. That is,
This is a highly convenient voice interaction device that smoothly performs a plurality of tasks.

The voice dialog device according to claim 2 is the voice dialog device according to claim 1, wherein the voice recognition unit is based on the estimation result of the work load estimation unit, and a plurality of recognition language dictionaries are provided. The optimal recognition language dictionary and the recognition acoustic dictionary are selected from the recognition acoustic dictionary and recognized. It is generally known that the utterances and pronunciations of workers vary depending on the work load. Therefore, the voice recognition unit selects the optimal recognition language dictionary and recognition acoustic dictionary corresponding to the estimation result (work load) of the work load estimation unit. Then, voice recognition is performed using the optimum dictionary. Since speech recognition is performed using the optimum dictionary, the utterance of the worker is always recognized accurately. As a result, the recognition rate is improved and a smooth dialogue can always be performed.

Further, the voice dialogue device according to claim 3 is the voice dialogue device according to claim 1, wherein the voice recognition unit detects the detection result of the work state detection unit and / or the dialogue progress degree information of the dialogue control unit. Based on the above, the optimal recognition language dictionary and the recognition acoustic dictionary are selected and recognized from the plurality of recognition language dictionaries and the plurality of recognition acoustic dictionaries. It is generally known that the acoustic characteristics and linguistic characteristics of a worker's utterance change depending on the work situation and noise situation. Therefore, the voice recognition unit prepares a recognition language dictionary and a recognition acoustic dictionary corresponding to various work situations and noise situations, and selects them according to the detection result of the work state detection unit. Thereby, the recognition rate can be improved. Also, the voice recognition unit obtains information necessary for recognition by feedback from the conversation progress degree information of the dialogue control unit. For example, if “Nagoya City” is confirmed in the address determination, the tail end of the detailed address becomes XX ward. The voice recognition unit does not need to recognize the last'ku '. That is, it is sufficient to recognize only the XX part. This can also improve the recognition rate. That is, the establishment of the voice dialogue can be improved.

Further, the voice dialogue apparatus according to claim 4 is the voice dialogue apparatus according to any one of claims 1 to 3, wherein the work state detecting section detects the worker state. The state detecting section and / or the apparatus state detecting section for detecting the state of the working apparatus and / or the noise detecting section for detecting ambient noise. The worker state detection unit directly detects, for example, a margin of the worker. The margin is detected from, for example, the heart rate, the posture change, and the number of limb movements. The device state detection unit directly detects the device state. When the device is a vehicle, the device state is, for example, the engine speed, the traveling speed, or the like. In addition, the noise detection unit detects the ambient noise level.
This is because the speech recognition rate of continuous words decreases when the noise level is high. If the work state detection unit is configured in this way, the voice recognition rate can be increased. Further, this makes it possible to further optimize the dialogue of the dialogue control unit at the latter stage of the voice recognition and to make the voice dialogue more accurate.

Further, the voice interactive apparatus according to a fifth aspect is the voice interactive apparatus according to the fourth aspect, wherein the worker state detecting section detects the worker state from the physiological index of the worker. The physiological index is a parameter indicating the physiological state of the worker. For example, the heart rate, the amount of sweat, etc. that change due to tension. This physiological index is a parameter that autonomously changes depending on the environment (work state), and the operator's margin, for example, can be accurately detected. If this is applied to the dictionary selection at the next speech recognition, the speech can be recognized more accurately. Further, if applied to the dialogue control unit via the work load estimation unit, work and voice dialogue can be performed more smoothly.

A voice dialogue system according to a sixth aspect of the present invention is the fifth aspect.
The voice interaction device according to the above item 1, wherein the physiological index of the worker is a heart rate, a sweat rate, and a line-of-sight-related value. These indexes are physiological indexes that allow the operator's margin and tension to be detected most easily. For example, when the degree of tension is high, the heart rate and the amount of sweating increase, and the viewpoint is fixed. The sensor for detecting the heart rate and the amount of sweating can be easily attached to the worker. Further, a simple monitoring camera for monitoring the line of sight can be easily attached to the working device. Therefore, if the heart rate, the amount of sweat, and the line-of-sight-related value are monitored by the respective sensors, it is possible to easily grasp the operator's margin and tension. Therefore, the voice dialogue apparatus according to claim 5 can be easily realized.

Further, the voice dialogue device according to claim 7 is the voice dialogue device according to any one of claims 4 to 6, wherein the device state detection unit includes the physical quantity of the work device and the device. The device status is detected from the frequency of use. Generally, there is a correlation between the worker status and the work equipment status. For example, if the working device is a vehicle, the margin of the driver who is a worker is low when traveling at high speed and high when traveling at low speed. Further, there is a correlation between the frequency of use of the device and the condition of the worker. For example, if the frequency of switch operations is high, the operator's margin is small. Therefore,
The worker state can be grasped from the working state of the working device. Therefore, the operator's state can be grasped more accurately. Therefore, it is possible to realize a voice dialogue device that makes work and dialogue compatible with each other more smoothly.

Further, the voice dialogue apparatus according to claim 8 is the voice dialogue apparatus according to any one of claims 1 to 7, wherein the dialogue progress information is confirmation items and confirmation of the confirmation items. One of the unknown state, the confirmation required state, and the confirmed state indicating the degree 1
It is composed of two. The voice dialogue apparatus of the present invention has dialogue progress degree information as described above. The conversation progress level information includes a confirmation item and a degree of confirmation of the confirmation item. And, the degree of certainty is unknown state (= 1), confirmation required state (=
2), which consists of three stages, the fixed state (= 3). For example, if the confirmation item is an address and it is in an unknown state, directly
Ask for an address. For example, it asks'Please tell me your address'. Also, if the confirmation is required, the question is implicitly asked at the time of the question of the next item. For example, in'Nagoya City '
What kind of restaurant is it? ' Then, when all the items to be inquired about are in a definite state, the process ends. As described above, the voice dialogue apparatus of the present invention advances the dialogue step by step using the dialogue progress degree information, and therefore does not reduce the work efficiency or stop the work. Further, the progress of the dialogue can be grasped and the dialogue can be surely performed.

Further, the voice dialogue apparatus according to claim 9 is the voice dialogue apparatus according to any one of claims 1 to 8, wherein the dialogue control unit has a small or medium workload. Uses a confirmation item in the confirmation required state to prompt confirmation of the item and sets a new question item. When the workload of the worker is small or medium, it is considered that the worker has room to confirm the recognition result of the previous question item and accept the new question item. For example, if the previous worker's response was'Piccolo in Nagoya City ',
If the new check item is an industry, 'What kind of shop is Piccolo in Nagoya? 'And set the question. This'Piccolo in Nagoya 'was the last item to check, and what kind of shop is it? 'Is a new question item. This repetition of the previous item will prompt the operator to implicitly confirm the previous item. That is, if there is no correction from the operator, the previous recognition was correct and if there is a correction,
Ask the question again. That is, if such a dialogue control unit is provided, a new question item can be set and an implicit understanding of the previous item can be obtained.

Further, the voice dialogue apparatus according to claim 10 is the voice dialogue apparatus according to any one of claims 1 to 8, wherein the dialogue control section has a large work load, The question items are set in a single question / answer format. When the workload of the worker is large, it is considered that the worker cannot afford to confirm the previous response item and accept the new question item. For example, if the previous worker's response was'Piccolo in Nagoya City 'and the new question item was the industry, regardless of the previous response item,' What kind of shop is it? What is'or 'industry? 'And the question item into one question and one answer form with a single sentence. Alternatively, before asking a new question, for example, confirm'Nagoya City 'and'Piccolo' explicitly. For example, 'Are you Nagoya? "Or" Piccolo? 'Is explicitly confirmed in a single sentence. That is, such a single sentence reduces the cognitive information processing load on the worker. That is, the work is continued and the dialogue is performed by reducing the cognitive information processing load of the worker. That is, even when the work load is large, work and voice conversation can be compatible.

[0027] Further, the speech dialogue apparatus according to claim 11 is the speech dialogue apparatus according to any one of claims 1 to 10, wherein the dialogue control section is used for a predetermined time and / or after a predetermined time. When the work load is estimated to be heavy, the interaction method, the interaction time interval and / or the utterance length, and the utterance speed are adjusted so that the interaction load does not increase for the predetermined time and / or after the predetermined time. For example, if the working device is a vehicle,
The work state detection unit may detect that the vehicle is accelerating. In such a case, it is considered that the worker (driver) has a heavy work load, especially a cognitive information processing load for a predetermined time until the speed change becomes constant. Also, for example, in a vehicle-mounted navigation device, it may be expected to turn left at an intersection after a predetermined time. In such a case, it is expected that the work load of the worker (driver) will be large after a predetermined time. In the present invention, when it is estimated that the predetermined time and / or the work load after the predetermined time is large, the conversation method, the conversation time interval and / or the utterance length, and the utterance speed are adjusted. For example, make the question item a simple sentence. Alternatively, for example, the conversation time interval is extended, or the utterance length and utterance speed of the question item are adjusted. That is, processing is performed so that the dialogue load does not increase for a predetermined time and / or after a predetermined time. Since the increase in work load does not overlap with the increase in conversation load, it is possible to perform work and conversation more smoothly. Note that adjusting the conversation time interval, adjusting the utterance length of the question item, and adjusting the utterance speed does not mean interruption of the conversation. The dialogue shall be carried out to the last item regardless of the time required.

[0028] Further, a voice interaction device according to a twelfth aspect is the voice interaction device according to any one of the first to eleventh aspects, wherein the work state detecting section and the work load estimating section are the voice recognizing section. The process is independent of the dialogue control unit and the guidance generation unit. Independent process means that there is no master-slave relationship. That is, since they operate independently of each other, for example, the voice recognition unit and the dialogue control unit can immediately obtain the work state information from the work load estimation unit at a necessary time. That is, the response speed can be increased. Therefore, it is possible to realize a voice dialogue device having an excellent response speed.

A voice dialogue system according to a thirteenth aspect is the voice dialogue system according to any one of the first to twelfth aspects, wherein the working device is a vehicle and the worker drives the vehicle. Person. The voice interaction device according to any one of claims 1 to 12 is characterized in that voice interaction is possible without disturbing the operation of the work device. Therefore, when the working device is a vehicle, it is possible to have a voice conversation without disturbing the driving of the vehicle. Therefore, it does not affect the safety of running the vehicle. That is, it becomes a voice dialogue device that guarantees safe driving.

Further, the voice dialog device according to claim 14 is the voice dialog device according to claim 13, wherein the voice dialog device indicates a physical quantity of the work device detected by the work state detection unit, that is, a vehicle state. It is meant that the parameter includes at least one of the following distance between the vehicle ahead and / or the vehicle behind, the position of the vehicle, the speed, the acceleration, the turning speed, the turning acceleration, and the accelerator / brake pedal depression amount. The inter-vehicle distance to the front vehicle and / or the rear vehicle detected by the work state detection unit represents the degree of approach of another vehicle. In this case, the dialogue method, speech length, etc. are adjusted according to the degree of proximity.
Further, the vehicle position represents, for example, the distinction between the driving lane and the overtaking lane. Since it is estimated that the driver's leeway varies depending on the lane, the conversation method, speech length, etc. are adjusted depending on the lane. When parameters such as speed, acceleration, and accelerator / brake pedal depression amount are used as parameters,
Or, it means decelerating. Since it is considered that the driver's leeway changes during acceleration or deceleration, the dialog system, utterance length, etc. of the voice dialog device are adjusted. Further, when the turning speed and the turning acceleration are used as parameters, it means that the vehicle is turning right or turning left, for example. Since it is considered that the driver's margin is small during the turning, the dialog system of the voice dialog device, the utterance length, etc. are adjusted in the same manner. In this way, if the vehicle state is discriminated using the physical quantity representing the vehicle state as a parameter, it becomes possible to perform voice interactive input while maintaining safe driving.

Further, a voice dialogue device according to a fifteenth aspect is the voice dialogue device according to the thirteenth or fourteenth aspects, and the voice dialogue device is provided in an information providing device mounted on a vehicle. The information providing device mounted on the vehicle is, for example, a navigation device, an audio device, a telephone device, or the like. If these devices are provided with the voice interaction device according to claim 13 or 14, information can be smoothly retrieved from the information providing device while maintaining safe driving of the vehicle. Therefore, the voice dialog device is excellent in vehicle safety and convenience that facilitates information acquisition.

[0032]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention will be described below. Some of the embodiments of the invention are described in the above-mentioned action and effect column. (First Embodiment) A voice interactive apparatus of the present invention will be described below with reference to the drawings. FIG. 1 shows an embodiment of a voice dialog device of the present invention. The voice interaction device of the present invention includes a voice input unit 10 for collecting voices of a worker, a voice recognition unit 12, a recognition language dictionary storage unit 11, a recognition acoustic dictionary storage unit 16, a database search unit 17, and a dialogue control unit 13. The guidance generation unit 14, the voice output unit 15, the work state detection unit 20, and the work load estimation unit 25. If this voice interaction device is applied to the information providing device 30 such as a navigation device, a voice interaction information providing device is formed. Although not shown, each component except the information providing device 30 is a CP.
U, ROM, RAM, A / D conversion device, D / A conversion device, system bus, external bus, external storage memory, sensor device, and computer device including programs in ROM.

Further, FIG. 2 shows the work state detecting section 20. The work state detection unit 20 includes a worker state detection unit 21 that detects a physiological index of the worker, a device state detection unit 22 that detects the driving state of the work device by a physical quantity, and noise detection that detects a noise environment around the worker. It is composed of the unit 23. The worker state detection unit 21 is, for example, an image processing device that monitors the motion of the worker and the movement of the pupil that is a line-of-sight related value. That is, it is a device that detects the margin based on the motion of the worker or the movement of the line of sight.
Then, the result is indicated by the signal A in the work load estimating unit 25 in FIG.
It is a device that outputs to. The device state detection unit 22 is a sensor device that detects the amount of depression of the accelerator / brake pedal when the working device is a vehicle, for example. Then, it is a device that outputs the detected amount as a signal B to the work load estimation unit 25. It should be noted that this sensor device may be a sensor device that detects not only the amount of depression of the accelerator / brake pedal but also the speed, acceleration, turning acceleration, etc. of the vehicle. Any sensor device may be used as long as it is a sensor device that detects a physical quantity that correlates with the operator's margin.

The noise detector 23 is, for example, a microphone device that detects noise around the worker, and outputs the result as a signal C to the work load estimator 25. Then, the work load estimation device 25 uses the signals A, B, C of the three parties.
This is a device that estimates the work load of the worker. For example,
When the accelerator / brake pedal depression amount is small, the work load is small. When the depression amount of any of the pedals is large, the work load is large. It should be noted that this detection and estimation is performed at predetermined time intervals, for example, by timer interruption.
Further, in the present embodiment, the working device is a vehicle, but this is one example, and the device may be any device operated by an operator. The voice interaction device of the present invention can be applied to various devices such as an aircraft, a ship, a lathe for production, and an automatic assembly machine for production, for example. The information providing device 30 is, for example, a navigation device, but this is also an example. In addition, the present invention can be applied to all information providing devices capable of providing various information by voice input of an operator such as an audio device, a computer device, and a central device of an automatic production machine.

Further, in FIG. 1, the voice input section 10 is, for example, a device including a microphone, an A / D converter and a memory (not shown). This is a device that detects a worker's voice with a microphone, converts it into digital data with an A / D converter, and stores it in a memory device. The voice recognition unit 12 extracts the feature amount of the voice digital signal sent from the voice input unit 10,
Recognition language dictionary storage unit 11, recognition acoustic dictionary storage unit 16
Is a device for identifying the input voice by referring to. At this time, the voice recognition unit 13 has a function of acquiring the work load from the work load estimation unit 25 and changing the recognition language dictionary and the recognition sound dictionary accordingly. This is because the utterance of the worker changes when there is a work load such as noise and a work load such as tension. The recognition rate is improved by changing the dictionary.

The dialogue control unit 13 is a device for searching a database that responds based on the recognition result of the voice recognition unit 12 and sequentially creating dialogue progress degree information based on the search result (for example, presence or absence of the database). is there. Here, the dialogue progress information is information including dialogue information and progress information. The dialogue information is composed of confirmation items or question items necessary for dialogue and parameters indicating the degree of confirmation thereof. For example, if the information providing device 30 is a navigation device, the facility name, address, detailed address, type of business, etc. of the destination are confirmation items, and the numerical value indicating the certainty thereof, for example, 1, 2, 3 is the degree of determination. For example, the degree of certainty 1 is an unknown state, the degree of certainty 2 is a confirmation required state, and the degree of certainty 3 is a confirmed state.

The progress information is the ratio of the above-mentioned confirmed state to all the confirmation items. For example, if the total number of confirmation items is 4 and the number of confirmed states is 2, the progress degree is 50.
%. The dialogue ends when the degree of progress is 100%. Further, the dialogue control unit sequentially advances the dialogue to create the dialogue progress degree information, and considers the result of the work load estimation unit 25 to provide the guidance generation unit 14 with necessary items and voice features for voice synthesis. It is a device for outputting. The guidance generation unit 14 is a device that creates a question response sentence (question guidance) or a confirmation response sentence (confirmation guidance) based on the item from the dialogue control unit 13. It is also a device for outputting the designated voice feature to the voice output unit 15 in the next stage. Further, it is a device that terminates when the progress level is 100% in the conversation progress level information and outputs, for example, the coordinates (latitude and longitude) of the destination to the target information providing device 30. The voice output unit 15 is a device that outputs the question response sentence or the confirmation response sentence input from the guidance generation unit 14 by using voice synthesis with a designated voice feature. The voice interaction device of this embodiment is configured in this way.

Next, the operation of the voice dialogue type information providing apparatus using the voice dialogue apparatus of this embodiment will be described with reference to the flowchart of FIG. 3 and the dialogue example of FIG. Here, the voice interactive information providing device will be described as a so-called in-vehicle navigation device. The operation of this device is divided into two, as shown in FIG. One is the process of the work state detection unit 20 shown in steps S30 to S32, and the other one is
One is the voice interaction process shown in steps S10 to S20. The process of the work state detection unit 20 is
For example, it is a program by a timer interrupt that is driven every 3 seconds to detect the work load. The process of the work state detection unit 20 first detects the work state in step S30. The work state is output as the above-mentioned worker state (signal A), device state (signal B), and noise state around the worker (signal C). Then, in step S31, those signals A,
The worker load is estimated from B and C. For example, the work load is classified into large, medium, and small, for example, by integrating the gaze change amount (signal A), the accelerator / brake depression amount (signal B), and the noise amount (signal C). Then, in step S32, the work load at that time is stored in the memory and the process ends. These workloads are quantities that change from moment to moment.

On the other hand, the voice dialogue process is step S10.
It starts from. First, in step S10, the above-mentioned dialogue progress degree information is initialized. For example, as shown in the uppermost column of FIG. 4B, the facility name (?, 1), address 1 (?, 1), address 2 (?, 1), industry (?, 1) are initialized. next,
Control goes to step S11. In step S11, the dialogue control unit 13 acquires the current work load from step S32. This is because the voice characteristics of the worker change depending on the work load, for example, the tension of the worker. Then, the process proceeds to step S12. In step S12, the speech recognition unit 12 selects a recognition language dictionary and a recognition acoustic dictionary according to the obtained work load. As a result, the voice recognition rate in the subsequent stage (step s18) is improved.

Then, the process proceeds to step S13. In step S13, the dialogue control unit 13 selects a question item and / or a confirmation item from the dialogue progress level information and the work load. At the time of initialization, as shown in the uppermost column of FIG. 4B, the dialogue progress level information is facility name (?, 1), address 1 (?, 1), address 2 (?, 1), industry (?, 1), the facility name (?, 1) and address 1 (?, 1) are selected, for example, and step S
Move to 14. In step S14, the guidance generation unit 14 receives the item and creates the first guidance, "Please tell me the name and address of the shop". Next, in step S15, it is output by voice synthesis (FIG. 4).
(A) Sys1). Note that FIG. 4A is an example of a dialogue obtained by this voice dialogue device, and FIG. 4B is a change diagram of dialogue progress degree information indicating the progress degree of the dialogue.

Then, the process proceeds to step S16. In step S16, it is determined whether or not the dialogue has ended. If not, the process proceeds to step S17. The method of determining the end will be described later. In step S17, the voice input from the driver is awaited. If there is a voice input, it is acquired and the process proceeds to step S18. In step S18,
The voice recognition unit 12 performs voice recognition using the recognition language dictionary and the recognition acoustic dictionary that are selected in step S12 and that match the work load. For example, driver's utterance (Drv1 :) 'um, piccolo in Nagoya. From ', recognize the facility name'Piccolo' and address 1'Nagoya City '. Then, the process proceeds to step S19.

In step S19, the dialogue control unit 13 confirms the presence or absence of the voice recognition result obtained in step S18 by searching a database, and updates the current dialogue progress degree information. That is, facility name (piccolo, 2), address 1 (Nagoya City,
2), address 2 (?, 1), and business type (?, 1) are created (Fig. 4 (b)). Then, the process returns to step S11 to repeat the subsequent routines. For example, step S13 and subsequent steps of the second routine will be taken as an example. Second step S1
In No. 3, for example, when the work load is small, it is considered that the driver's margin is large, and since the facility name and address 1 are in the confirmation required state 2 in the first routine, it is again implicitly ' Select “City of Nagoya” and “Piccolo” as confirmation items, and select other items, such as industry, as question items. In other words, “Nagoya City” and “Piccolo” that are in the confirmation required state and the unconfirmed state 1
“Industry” is selected and the process proceeds to step S14.

In step S14 for the second time, guidance is generated according to the selected item. For example, if'Nagoya City 'and'Piccolo' are items to be checked and the industry item is a question item, what kind of store is'Piccolo in Nagoya city '? Set the question and answer sentence of '(Sys2 :). Then, in the next step S15, it is output to the driver by voice synthesis of the voice feature specified. In this way, the dialogue is sequentially repeated and the dialogue progress information is updated. That is, FIG.
The new dialogues shown after Drv2 in (a) are repeated, and finally the progress of the dialogue progress information in FIG. 4 (b) becomes 100%.

The method of ending is determined from this dialogue progress degree information. That is, if the degree of determination of each item of the dialogue progress degree information is all 3, it is judged that the dialogue has ended. That is, the degree of progress 10
If it is 0%, the process ends. That is, at this time, the dialogue progress information is the facility name (piccolo, 3), address 1 (Nagoya City,
3), address 2 (Naka Ward, 3), type of business (restaurant, 3). Then, the process proceeds to the final step S20. In step S20, the converted data (latitude, longitude) of'Restaurant Piccolo 'is output to the information providing device 30 which is a navigation device, for example. Then, the information providing device 30 '
For example, map information of the restaurant Piccolo 'is output to a display device (not shown). The voice interactive information providing apparatus using the voice interactive apparatus of the present invention operates in this way.

As described above, in the present embodiment, the recognition language dictionary and the recognition acoustic dictionary are selected according to the work state, so that the voice recognition can be accurately performed. Furthermore, since the dialogue control unit changes the dialogue method depending on the work state, the worker does not stop the operation of the work device or reduce the work efficiency. That is, the work and the voice conversation can be compatible with each other. In particular, when the present invention is applied to a vehicle as described above, it is possible to enable voice interactive input and safely maintain the driving operation of the driver. That is, it is possible to enable interactive input while maintaining safe driving.

In step S13 of the second time, the workload is set to be small, "Nagoya City" and "Piccolo" are required confirmation items, and the business type item is another question item. Is it such a store? The question and answer sentence "is set, which is created from the correspondence table between the workload estimation amount and the dialogue control method shown in FIG. Create for other workloads from this correspondence table. The correspondence is briefly described below. (1) If the work load estimation amount is small, the interactive control method includes a plurality of question items, and if there is a confirmation required state, an implicit confirmation item and a single question item. For example, say the'name 'and'address' of the store (multiple question items). If you have a status that requires confirmation, please check the Nagoya city's'
What kind of shop (industry) is Piccolo? (Implicit confirmation item + multiple or singular question item). (2) If the work load is medium, the interactive control method uses a single question item, and if the user has a confirmation-required state, an implicit confirmation item and a single question item. For example, say the'name 'of your store (single question item). Also, if you have a status that requires confirmation, what kind of shop (facility name) in'Nagoya City '? (Implicit confirmation item + single question item). (3) When the work load is large, the interactive control method is a single question item, and when it has a confirmation required state, it is an explicit confirmation item. For example, say the'name 'of your store (single question item). Or is it'Nagoya City '?
(Explicit question item). In step S13, FIG.
Items are selected by the correspondence table shown in FIG. 3 and the interactive control method is determined accordingly. The slots in the figure are used to mean items.

(Second Embodiment) In the first embodiment, the voice recognition unit 12 also uses the work load by the work load estimation unit 25 as a parameter to select the recognition language dictionary and / or the recognition sound dictionary. . Alternatively, the voice recognition unit 12 may feed back the dialogue progress degree information created by the dialogue control unit 13 in the next stage and use it as a parameter.
That is, the configuration shown in FIG.
It may be fed back to the voice recognition unit 12. For example, in the dialogue progress level information of the first embodiment, there is an example of address 1 (Nagoya city, 3) and address 2 (?, 1). That is, if the city of Nagoya at the destination address is in the confirmed state 3,
It is self-evident that the item of address 2 is always XX ward. That is, in the voice recognition, the last pronunciation is'ku '.
That is, it is not necessary to recognize the last pronunciation. That is, the tail end is limited to “ku (district)”, and feedback is performed so as to recognize the rest. This improves the recognition rate and the recognition speed. Also, for example, industry (restaurant, 3),
When the address becomes 1 (Nagoya city, 3) and the address 2 (Naka-ku, 3), the voice recognition unit 13 can refer to the Naka-ku restaurant database in recognizing the facility name. The recognition result can be the store name of the utterance that matches or is closest to the restaurant database in Naka Ward. That is, the recognition rate can be improved also in this case.

(Modification) The above embodiment is only one example, and various modifications can be considered. For example, in the first embodiment, one of the work loads is determined from the accelerator / brake pedal depression amount, but the work load at that time is determined.
Even if the accelerator is turned off from N, it is in a high-speed state for several seconds, and as shown in FIG. 7, the work load is expected to be large (arrow). In such a case, even if the work load at that time is medium, it is largely corrected, and the interval between the question response Sys2 and the confirmation response Sys3 from the system is set longer than usual.
You may do this. That is, it is set so that the work load (driver's tension) and the response load (driver's cognitive information processing load) do not overlap, that is, the dialogue load does not increase. As a result, the load on the driver can be reduced and safe driving can be maintained.

Further, the worker state detecting section 22 of the first embodiment.
The image processing device is an image processing device that monitors the movement of the pupil, which is a line-of-sight-related value of the driver, as a physiological index of the worker.
The same effect can be obtained. The point is that a sensor device that can obtain a correlation with the operator's margin or tension is sufficient. Further, although the apparatus state detection unit 23 of the first embodiment sets the detected physical quantity as the accelerator / brake pedal depression amount of the vehicle, it may be the depression frequency. The point is that it may be a parameter of the device (vehicle) that is correlated with the driver's load.

Although the order of selecting items is not mentioned in the first and second embodiments, the order is arbitrary. For example, in the dialogue example of FIG. 7, it is determined in the order of address 1, address 2, facility name, and business type, but in the dialogue example of FIG. 8, it is determined in the order of address 1, facility name, address 2, and business type. There is. The point is that the dialogue progress level information will eventually reach 100%. The order does not matter.

Further, in the first embodiment, the voice recognition unit 12 is provided with a plurality of recognition language dictionaries and a recognition acoustic dictionary, and the optimum dictionary is selected from them based on the estimation result of the work load estimation unit 25. However, the voice recognition unit 12 may directly refer to the detection result of the work state detection unit 20 and select the optimum dictionary based on the result (FIG. 9). For example, the ambient noise level in the surroundings or the worker's state may be directly detected from the work state detecting unit 20, and the dictionary corresponding thereto may be selected. Since the work status can be grasped directly, a more accurate dictionary can be selected. Further, with the system configuration as shown in FIG. 9, the processing by the work load estimating unit 25 is bypassed, so that the recognition language dictionary and the recognition acoustic dictionary can be quickly selected. At this time, step s12 in FIG.
In the above, the voice recognition unit 12 refers not to the work load of the work load estimation unit 25 but to the detection result of the work state detection unit 20.

Further, in the first embodiment, the voice recognition unit 12 includes a plurality of recognition language dictionaries and a recognition acoustic dictionary, and selects an optimum dictionary from them based on the estimation result of the work load estimation unit 25. However, if the work environment is always the same and the workers are always the same, the dictionary may be singular. That is, the voice recognition unit 12 may always recognize the same dictionary without referring to the estimation result of the work load estimation unit 25 (FIG. 10). The same effect can be obtained. In this case, step s12 in FIG. 3 is omitted.

[Brief description of drawings]

FIG. 1 is a system block diagram of a voice dialog device according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram of a work state detection unit according to the first embodiment of the present invention.

FIG. 3 is a flowchart showing the operation of the voice dialog device according to the first embodiment of the present invention.

FIG. 4 is a diagram showing a dialogue example (a) by the voice dialogue apparatus according to the first embodiment of the present invention and a change diagram (b) of dialogue progress degree information due to the dialogue.

FIG. 5 is a correspondence table between a workload estimation amount and a dialogue control method according to the voice dialogue apparatus of the first embodiment of the present invention.

FIG. 6 is a system block diagram of a voice dialog device according to a second embodiment of the present invention.

FIG. 7 is a modification of the dialogue example according to the first embodiment of the present invention.

FIG. 8 is a modification of the dialog example according to the first embodiment of the present invention.

FIG. 9 is a system block diagram of a voice interaction device according to a modification of the first embodiment of the present invention.

FIG. 10 is a system block diagram of a voice interaction device according to a modification of the first embodiment of the present invention.

FIG. 11 is a system block diagram of a conventional voice dialog device.

[Explanation of symbols]

10 ... Voice input section 11 ... Recognition language dictionary storage unit 12 ... Voice recognition unit 13 ... Dialogue control unit 14 ... Guidance generator 15 ... Voice output section 16 ... Acoustic dictionary storage for recognition 17 ... Database search section 18 ... Database 20 ... Working state detector 21 ... Worker state detection unit 22 ... Device state detector 23 ... Noise detector 30 ... Information providing device

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Noriyoshi Terazawa             Aichi Prefecture Nagachite Town Aichi District             Ground 1 Toyota Central Research Institute Co., Ltd. (72) Inventor Hiroyuki Hoshino             Aichi Prefecture Nagachite Town Aichi District             Ground 1 Toyota Central Research Institute Co., Ltd. (72) Inventor Toshihiro Wakita             Aichi Prefecture Nagachite Town Aichi District             Ground 1 Toyota Central Research Institute Co., Ltd. F term (reference) 5D015 HH06 HH13 KK02 LL10

Claims (15)

[Claims] 1. A voice interactive device that enables voice input during work by using voice recognition technology, comprising: a work state detection unit; and a work performed by a worker based on a detection result of the work state detection unit. A workload estimation unit for estimating a load; a speech recognition unit for recognizing an input from a speech input unit by using one or more recognition language dictionaries and one or more recognition acoustic dictionaries; and the speech recognition unit. Of the recognition result, a database search result corresponding to the recognition result, dialogue progress degree information composed of the database search result and the degree of confirmation, and the work load level estimated by the work load estimating unit, A dialogue control unit that selects a confirmation item and / or question item and a voice feature that does not increase the workload of the worker, and a confirmation guidance and / or question guider for the worker based on the result of the dialogue control unit. Voice dialogue system, characterized in that it includes a guidance generator for generating a scan. 2. The voice recognition unit is configured to optimize the recognition language dictionary and the recognition language dictionary from the plurality of recognition language dictionaries and the plurality of recognition acoustic dictionaries based on the estimation result of the work load estimation unit. The voice interaction device according to claim 1, wherein a voice dictionary is selected and recognized. 3. The voice recognition unit, based on the detection result of the work state detection unit and / or the dialogue progress degree information of the dialogue control unit, the plurality of recognition language dictionaries and the plurality of recognition language dictionaries. 2. The voice dialogue apparatus according to claim 1, wherein the optimal recognition language dictionary and the recognition acoustic dictionary are selected and recognized from the acoustic dictionary. 4. The work state detecting section includes a worker state detecting section for detecting a worker state, an apparatus state detecting section for detecting a state of a working apparatus, and / or a noise detecting section for detecting ambient noise. The voice interaction device according to any one of claims 1 to 3, wherein: 5. The voice interaction device according to claim 4, wherein the worker state detecting unit detects the worker state from a physiological index of the worker. 6. The voice interactive apparatus according to claim 5, wherein the physiological index of the worker is a heart rate, a sweat rate, and a line-of-sight related value. 7. The apparatus state detecting unit according to claim 4, wherein the apparatus state detecting unit detects the apparatus state based on a physical quantity of the working apparatus and an operation frequency of the apparatus. Spoken dialogue device. 8. The dialogue progress level information is composed of a confirmation item and one of an unknown state, a confirmation required state, and a confirmation state indicating the degree of confirmation of the confirmation item.
9. The voice interaction device according to claim 7. 9. The dialogue control unit, when the workload is small or medium, prompts confirmation of the item by using a confirmation item in a confirmation required state and sets a new question item. The voice interaction device according to any one of claims 1 to 8. 10. The dialogue control unit sets a question item in a single question-and-answer form when the workload is large, according to any one of claims 1 to 8. The voice interaction device according to item 1. 11. The dialogue control unit prevents the dialogue load from increasing when the predetermined time and / or the work load after the predetermined time is estimated to be large. The dialogue method, the dialogue time interval, and / or the speech length and the speech speed are adjusted.
11. The voice interaction device according to claim 10. 12. The work state detection unit and the work load estimation unit are processes independent of the voice recognition unit, the dialogue control unit, and the guidance generation unit. 11. The voice interaction device according to any one of 11. 13. The voice interactive apparatus according to claim 1, wherein the working device is a vehicle, and the worker is a driver of the vehicle. 14. The physical quantity of the working device includes at least one of a following distance between a front vehicle and / or a rear vehicle, a position of the vehicle, a speed, an acceleration, a turning speed, a turning acceleration, and an accelerator / brake pedal depression amount. 14. The voice interaction device according to claim 13, wherein: 15. The voice interactive apparatus according to claim 13, wherein the voice interactive apparatus is provided in an information providing device mounted on the vehicle.