JP2000020086A - Speech recognition apparatus, navigation system using this aperture and vending system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a speech recognition apparatus which prevents erroneous recognition as far as possible and is improved in operability. SOLUTION: For example, the request by a user in the state that a restaurant list is displayed may be either the assignment of the store name of the restaurant itself or the genre assignment to discriminate the genre of the restaurant further in details. Then, a recognition control section 35 acquires the state that the restaurant list is displayed from an operation control section 5. This control section, them, assigns the 'dictionary relating to the restaurant' to a dictionary management section 34. This dictionary management section 34 edits the selected dictionary extracting necessary component in accordance with the dictionary assignment from a pre-stored base dictionary and sets this dictionary in a recognition processing section 33. The five items; the prefectures, the cities, towns and villages, genre, details and conditions, are set in the base dictionary and, therefore, only the recognition vocabulary having the genre of a 'meal' is extracted in accordance with the genre among these items and the selected dictionary is set in this case.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech recognition device effective for, for example, a case where a destination setting or the like in a navigation system can be input by voice, a navigation system equipped with the speech recognition device, and a vending system. About.

[0002]

2. Description of the Related Art Conventionally, there has been proposed a speech recognition apparatus which compares an inputted speech with a plurality of pattern candidates to be compared which are stored in advance and determines a speech having a high degree of coincidence as a recognition result. It has already been put into practical use, and is used, for example, for a user to input a destination name to be set in a navigation system by voice. In particular, when the driver uses the in-vehicle navigation system, voice input does not involve button operation or screen gaze, so that it is effective because the safety is high even when the vehicle is running.

[0003] However, in the current recognition technology, the recognition result is not always correct, and it is inevitable that some degree of erroneous recognition will occur. Therefore, when erroneous recognition occurs, it is necessary to redo speech recognition itself, and it may be necessary to take considerable time and effort to obtain a result desired by the user. From such a point, a device for minimizing erroneous recognition is expected.

Accordingly, an object of the present invention is to provide a speech recognition apparatus which solves such a problem and prevents erroneous recognition as much as possible, thereby improving usability. It is another object of the present invention to provide a navigation system and an automatic vending system using the voice recognition device.

[0005]

According to the first aspect of the present invention, there is provided a speech recognition apparatus which, when a user inputs a speech through a speech input means, recognizes the speech. The means compares the input speech with a plurality of pattern candidates to be compared previously stored in the dictionary means, and determines a speech having a high degree of coincidence as a recognition result. At this time, the dictionary management means performs the following management on the dictionary means. That is, a state in the external device that handles the recognition result as the instruction content of the user is input, and based on the input state, only a predetermined one of the plurality of comparison target pattern candidates stored in the dictionary unit is temporarily used. And only the extracted comparison target pattern candidates are validated as dictionary data for recognition by the recognition means.

One of the causes of the erroneous recognition in the conventional speech recognition described above is the contents of dictionary data used for recognition. In other words, in the voice recognition, the input voice is compared with a comparison target pattern candidate stored in advance, and a voice with a high degree of matching is used as a recognition result.
Since the comparison target pattern candidates are similar to each other, a pattern different from the desired comparison target pattern is determined as a recognition result. In this regard, in the voice recognition device of the present invention, even if the comparison target pattern candidate is set in the dictionary means in advance, the comparison target pattern is not validated as the dictionary data at the time of recognition by the dictionary management means. Pattern candidates are not compared at the time of recognition, and are naturally not recognized as recognition results. By narrowing down the dictionary data in this manner, occurrence of erroneous recognition can be prevented as much as possible.

[0007] The dictionary data is not simply narrowed down, but is narrowed down based on the state of an external device that handles the recognition result as the content of a user's instruction. For example, as set forth in claim 2, the dictionary management means invalidates the comparison target pattern which is considered impossible as an instruction from the user in the input state of the external device as dictionary data for recognition. By doing so, it is conceivable to make other things valid. In this way, the comparison target pattern considered necessary from the state of the external device at that time remains valid as it is, and the other patterns become invalid as dictionary data. Speech recognition for the vocabulary to be performed can be appropriately realized.

[0008] In the above speech recognition device,
Since only a part of the dictionary data is valid, the invalidated dictionary data is not compared at all at the time of recognition. Therefore, there is a concept that a vocabulary that can be an instruction is effective as dictionary data even though the possibility of the instruction from the user is low.
However, if the data having such a low possibility is made effective as the dictionary data, the vocabulary which may be misrecognized in the dictionary data eventually increases. In view of this point, it is also preferable to employ the voice recognition device according to the third aspect.

In the case of the voice recognition device of the third aspect, when the user inputs voice through the voice input means, the recognition means transmits the input voice to a plurality of comparison means stored in the dictionary means in advance. Those having a higher matching degree than the target pattern candidates are regarded as recognition results. At this time, the dictionary management means performs the following management on the dictionary means. That is, a state in the external device that handles the recognition result as the instruction content of the user is input, and based on the input state, the priority is temporarily set for a plurality of comparison target pattern candidates stored in the dictionary unit, The higher the priority of the set comparison target pattern candidate, the easier it is to use as dictionary data for recognition by the recognition means. For example, as set forth in claim 4, it is possible to make the comparison target pattern considered to have a higher probability of being possible as an instruction from the user in the state of the input external device easier to be used as dictionary data for recognition. Conceivable.

In this way, for example, even if there are two comparison target pattern candidates that are likely to be erroneously recognized, one is set to have a high priority, and if the other is set to be low, the priority is set to be low. The possibility that the comparison target pattern candidate is erroneously recognized is reduced, and as a result, erroneous recognition can be prevented as much as possible. Also, a comparison target pattern candidate set with a low priority has a low priority but finally exists as a recognition target, so that it can be obtained as a recognition result. Therefore, for example, the speech recognition device outputs the recognition results in the order of the comparison target pattern candidates with the highest priority, but if the user instructs to output another recognition result because the recognition result is not desired,
Eventually, the comparison target pattern candidate having a lower priority is output as a recognition result, which can be desired by the user.

In other words, in a situation where it is impossible to cope with the situation by completely using the option of using or not using, the method of setting the priority is effective. By the way, when the voice recognition device according to any one of claims 1 to 4 is used for a navigation system, it is conceivable to configure as shown in claim 5. In other words, the voice recognition device according to any one of claims 1 to 4 and a navigation device are provided, and the voice input means of the voice recognition device needs to be specified at least when the navigation device performs a navigation process. This is used for the user to input an instruction of predetermined navigation processing-related data by voice.

[0012] In this case, the "predetermined navigation processing-related data" is a typical destination, and the destination may be the place name itself, or a facility or store name. In some cases. Also,
In addition, instructions that need to be specified in performing the navigation process, such as selecting conditions for route search, are included.
When a destination is designated by a facility, a store name, or the like, it is often the case that conditions are set from a database and narrowed down.
In that case, for example, when narrowing down by a genre such as a restaurant, dictionary data about a facility or a store name that cannot be provided as a meal providing place may be invalidated or set to a low priority. In particular, since a general navigation device can display a map or a menu on a display device, the dictionary management means can use the display content based on the content displayed on the display device. It is only necessary to invalidate the comparison target pattern that is considered impossible as an instruction from the user as dictionary data for recognition or to lower the priority. In the above example, if the screen display shows a list of restaurants, it is not possible to request a map operation such as "change map orientation" or "change map scale" in that state. The candidate for the comparison target pattern of the vocabulary related thereto may be invalid or have a low priority.

On the other hand, when the speech recognition device according to any one of claims 1 to 4 is used for an automatic vending system, it is conceivable that the speech recognition device is configured as shown in claim 7.
In other words, the apparatus comprises the voice recognition device according to any one of claims 1 to 4 and a vending device, and is provided with a voice input unit of the voice recognition device, at least when the vending device processes a product to be sold. This is used by a user to input an instruction of predetermined vending process-related data which needs to be specified by the user.

The automatic vending apparatus referred to here is, for example, a vending apparatus for cigarettes, drinking water, and the like. When a user specifies the brand of the product by voice, the specified product is sold. . In this case, for example, as described in claim 8, the voice recognition device uses comparison target pattern candidates relating to proper nouns of all items theoretically available for sale by the vending device as dictionary data in the dictionary means. In addition to the above, the dictionary management means temporarily disables, as dictionary data, comparison target pattern candidates relating to proper nouns other than articles that can be actually sold at the time of the vending device, or sets the priority. It is conceivable to lower it. For example, in the case of a cigarette vending system, comparison target pattern candidates corresponding to the proper nouns of cigarettes of all brands theoretically available for sale are prepared as dictionary data. Then, for each vending device, for comparison pattern candidates relating to proper nouns other than the brand of cigarettes actually sold by the device, temporarily invalidate them as dictionary data or set a low priority. deep. Specifically, the administrator of the vending apparatus may manually set. In this way, the dictionary data can be adjusted according to what brand of cigarette is selected and subject to vending, so that a cigarette of a brand not handled by the vending machine is regarded as a misrecognition result. No more. Therefore, the convenience for the user of the vending machine is improved. In this case, regarding the comparison target pattern candidates related to the proper nouns of the brands not handled,
It can be said that invalidating the dictionary data is preferable to lowering the priority.

[0015]

Embodiments of the present invention will be described below with reference to the drawings. It is needless to say that the embodiments of the present invention are not limited to the following examples, and can take various forms as long as they belong to the technical scope of the present invention.

[First Embodiment] FIG. 1 is a block diagram showing a schematic configuration of an on-vehicle navigation system 1 to which a voice recognition device 30 is applied. The present navigation system 1
Is an operation switch unit 4 including a voice recognition device 30, a mechanism switch 2, and a remote control switch 3, an operation control unit 5 for inputting operation status and the like from the operation switch unit 4 and controlling the entire navigation system, Current position calculating unit 6 for calculating the current position, an appropriate route to the destination based on the destination set by the user (recommended route)
Based on the recommended route calculated by the route calculation unit 7, a control process for providing the user with voice or a simple map or the like necessary for smoothly driving the vehicle is performed. It includes a guidance control unit 8, a map data search unit 9, an audio output control unit 11, a speaker 12, a display control unit 13, and a display 14. The operation control unit 5 is configured as a normal computer, and includes a well-known CPU, ROM, RAM, I / O, and a bus line for connecting these components.

The current position calculating unit 6 detects a vehicle position based on radio waves from a well-known gyroscope, a distance sensor, and a satellite, for example.
), and a GPS receiver for the itioning system). Since each of these sensors and the like has an error having a different property, a plurality of sensors are used while interpolating each other. It should be noted that depending on the accuracy, it may be constituted by a part of the above, and furthermore, a geomagnetic sensor, a rotation sensor for steering, a wheel sensor for each rolling wheel, or the like may be used.

The map data search section 9 is a device for inputting various data including so-called map matching data, map data and landmark data for improving the accuracy of position detection. As a medium, CD-R
Generally, an OM or a DVD is used, but another medium such as a memory card may be used.

The display 14 is a color display.
The vehicle current position mark input from the map data, the map data input from the map data search unit 9, and additional data such as a guide route to be displayed on the map and a mark of a set point to be described later are displayed in a superimposed manner. it can.

As the mechanism switch 2, for example, a touch switch or a mechanical switch integrated with the display 14 is used, and is used for a destination for route calculation and various inputs. Further, the same instruction input as that of the mechanism switch 2 can be input via the remote control switch 3. The voice recognition device 30 is used when the mechanism switch 2 or the remote control switch 3 is used to manually specify a destination or the like. Is a device for enabling the user to instruct the user. Note that the operation switch section 4 does not necessarily have to include all of the voice recognition device 30, the mechanism switch 2, and the remote control switch 3, and any one of them can input a desired instruction. However, in the present embodiment, it is assumed that three are provided for convenience.

Next, the speech recognition device 30 constituting a part of the operation switch section 4 will be described in more detail with reference to FIG. The voice recognition device 30 includes a microphone 31 as “voice input means”, an input signal processing unit 32 that converts a voice signal input via the microphone 31 into a form capable of data processing by A / D conversion or the like, and “ A recognition processing unit 33 as a “recognition unit”, a dictionary management unit 34 as a “dictionary unit” and a “dictionary management unit”, and a recognition control unit 35 also functioning as a part of the “dictionary management unit” are provided. .

Although not shown in FIG. 2, the voice input method according to the present embodiment uses a PTT (Push-To-
Talk) inputting a voice via the microphone 31 while pressing the switch. Specifically, the input signal processing unit 32 determines whether or not the PTT switch has been pressed. If the PTT switch has been pressed, the input signal processing unit 32 executes a voice input process via the microphone 31. If not, the voice input process is not executed. Therefore, only the voice data input via the microphone 31 while the PTT switch is pressed is output to the recognition processing unit 33.

The recognition processing unit 33 compares the voice data input from the input signal processing unit 32 with the comparison target pattern candidates in the set dictionary data, and determines a higher comparison target pattern having a high degree of coincidence with the recognition control unit 35. Output to The recognition of the word sequence in the input speech is performed by sequentially performing acoustic analysis on the speech data input from the input signal processing unit 32 to extract an acoustic feature (for example, cepstrum), and the acoustic feature obtained by the acoustic analysis. Get time series data. And the well-known D
This time series data is divided into several sections by the P matching method, and it is determined which word each section corresponds to stored in the dictionary data.

It should be noted that the recognition processing section 33 sends the recognition control section 35
Is not necessarily one. For example, there may be a plurality of vocabulary words ranked higher in the similarity determination result in the recognition processing unit 33 or a plurality of recognized vocabulary words. Since the voice recognition technology is a pattern matching between a registered recognition target word and a fetched voice, it is common to output a plurality of words having a high degree of matching as a recognition result. For example, when a prefecture name is used as a recognition dictionary, if the patterns of “Wakayamaken” and “Okayamaken” are similar, the above two may be output as a recognition result in response to the utterance of the user “Okayama Prefecture”. . Also, in recent technology, user utterances, such as “Nagoya City Art Museum”
It is also possible to recognize the two vocabularies of “Museum” and “Museum” as separate things and output the recognition result.

Then, the recognition control unit 35 outputs the recognition result to the operation control unit 5 and acquires the state from the operation control unit 5. This state acquires the current setting state of the navigation system 1 and the like. Then, a dictionary is specified to the dictionary management unit 34 based on the acquired state.

The dictionary management unit 34 sets the stored dictionary data in the recognition processing unit 33. The recognition processing unit 33 performs the recognition process based on the set dictionary data. However, the dictionary management unit 34 does not always set the stored dictionary data in the recognition processing unit 33 in the same state. That is, when a dictionary is specified from the recognition control unit 35, the dictionary data edited based on the specification is set in the recognition processing unit 33. The contents of this dictionary editing will be described later.

Next, the operation of the navigation system 1 of this embodiment will be described. In addition, since the part related to the voice recognition device 30 is a feature, the general operation of the navigation system will be briefly described, and then the operation of the part related to the voice recognition device 30 will be described in detail.

After the power of the navigation system 1 is turned on, the driver selects the route information display processing from the menu displayed on the display 14 by using the mechanism switch 2 or the remote control switch 3 to display the guide route on the display 14. Alternatively, by inputting a desired menu by voice through the microphone 31 through the voice recognition device 30, the same instruction as that selected through the switches 2 and 3 described above is operated and controlled by the recognition control unit 35. When the processing is performed on the unit 5, the following processing is performed.

That is, the driver operates the display 14
When the destination is input by voice or switch operation based on the above map, the current position of the vehicle is obtained by the current position calculation unit 6, and the cost is calculated between the destination and the current position by the Dijkstra method. A process of obtaining the shortest route from the route to the destination as a guidance route is performed. Then, the guidance route is displayed on the road map on the display 14 to guide the driver to an appropriate route. The calculation process and the guidance process for obtaining such a guide route are generally well-known processes, and a description thereof will be omitted.

Next, the operation of the speech recognition apparatus 30 will be described with reference to the flowchart of FIG. In the first step S10 upon activation, it is determined whether or not there is a user setting. The user settings here are
For example, it is a setting of a point to be specially targeted for voice recognition (in the navigation system, point information registered as something other than an address place name and a facility name as a memory point registration), a method of confirming a voice recognition result, and the like. If such a user setting has not been made (S10: NO),
The process directly proceeds to S30, but if the user setting has been made (S10: YES), the setting state is input (S20), and then the process proceeds to S30.

In step S30, the state of the navigation system, for example, a state such as "the map currently displayed on the display 14 is a 1 / 10000th scale road map, and there is no more detailed map" is fetched. As a display other than a map, for example, when searching for destinations by genre, there may be a state where a "restaurant list", a "hot spring list", a "golf course rust", etc. are displayed. Captures whether the list is displayed.

The input of the setting state in S20 and the input of the system state in S30 are performed by the recognition control unit 35 shown in FIG. Then, in subsequent S40, the recognition control unit 35 determines that S2
Based on the setting status input at 0 and the system status fetched at S30, a dictionary is specified to the dictionary management unit 34. Then, the dictionary management unit 34 sets a dictionary in the recognition processing unit 33 based on the dictionary specification. For example, the dictionary data (base dictionary) shown in FIG. 5 includes all recognizable vocabularies, and only predetermined data is extracted from the dictionary data (selection dictionary) shown in FIG. It is set. The specific setting will be described later.

When the preparation for speech recognition is completed, speech recognition is performed in S60, and it is determined whether or not the speech recognition has succeeded. The voice recognition is performed as follows. As described above, only the audio data input via the microphone 31 while the PTT switch (not shown) is being pressed is output from the input signal processing unit 32 to the recognition processing unit 33. Therefore, if voice data is input from the input signal processing unit 32, the recognition processing unit 33
The collation is performed by using the dictionary data set in the above, and the recognition process is executed. And there is an abnormality in the capture of audio data,
Alternatively, as a result of the collation using the dictionary data, if a defect such as no match is detected, it is determined that the voice recognition has failed (S60: NO), and the present processing routine is terminated as it is.

On the other hand, if the speech recognition is successful (S6
0: YES), the recognition control unit 35 outputs the recognition result to the operation control unit 5. Then, the recognition result is voice-responded from the operation control unit 5 via the voice output control unit 11 and the speaker 12 (S70). By the voice response of the recognition result, the user can determine whether or not the instruction intended by the user is correctly recognized.
For example, by inputting "Yes" by voice or operating a predetermined confirmation switch, the user request is confirmed (S80: YES), and the process proceeds to S90. If the user request is not confirmed, the process returns to S60 to perform the recognition process on the input voice data again. Also,
The determination of the user request may be made by providing a switch for instructing the determination, for example, in the mechanism switch 2 or the remote control switch 3 and operating the switch.

In S90, the recognition control section 35 responds to the operation control section 5 with the request confirmation. With this, for example, when a map at a scale of 1/10000 is displayed on the display 14, and a voice input from the user is a request to "change the map scale to 1/20000", When the request is determined, the operation control unit 5 executes a process of changing the scale of the map displayed on the display 14 to 1/20000. Further, in the case of a user request such as "set a destination", the request cannot be realized without the point information as the destination. Therefore, in this case, in S20 in the process of FIG. 3 to be executed again, it is input that the user setting is a state of “setting a destination”, and based on the setting state, the destination is set in S40 and S50. Performs dictionary editing and setting for setting. Then, the voice recognition processing is executed. In this manner, the function requested by the user is
Repeat the voice recognition process until realized by.

In FIG. 3, the overall operation of the speech recognition apparatus 30 has been described. However, since the dictionary editing and setting in S40 and S50 are characterized, this point will be further described with specific examples. For example, when the voice recognition function is activated while the list of “restaurants” is displayed on the display 14 as shown in FIG. 4A, the restaurant list is displayed in response to a user operation. Obviously, in this state, it is unlikely that the user makes a request for a map operation such as "change of map orientation" or "change of map scale". Also, it is unlikely that other lists, for example, data on hot springs and golf courses required when the “hot spring list” and “golf course list” are displayed, are requested in this state.

Therefore, the user requests while the restaurant list is displayed as shown in FIG. 4A, either by specifying the restaurant name itself, or by specifying a genre for discriminating the genre of the restaurant in more detail. It can be assumed to be one of them, and can be narrowed down. Therefore, the recognition control unit 35 in FIG. 2 obtains a state that the restaurant list is displayed from the operation control unit 5 and specifies the “dictionary about restaurant” to the dictionary management unit 34.

The dictionary management unit 34 edits a selected dictionary (see FIG. 6) in which necessary parts are extracted (or unnecessary parts are deleted) based on dictionary designation from a base dictionary (see FIG. 5) stored in advance. , In the recognition processing unit 33. In the base dictionary shown in FIG. 5, five items of prefecture (however, Nagoya City is included here), municipalities, genre, details, and conditions are set for each recognized vocabulary. In this case, Based on the genre, only the recognition vocabulary having the genre “meal” is extracted, and the selection dictionary shown in FIG. 6 is set. When the selection dictionary is set in this way, for example, in FIG. 5, "OO taste", "Izumi town", "Izumi city"
There are three vocabulary words, "Izumi-cho" and "Izumi-shi", which have a high possibility of misrecognition for "oo taste".
The selected dictionary (FIG. 6) from which the recognized vocabulary is deleted can be set. Therefore, when the user wants to specify “XX taste”, the “XX taste” is appropriately recognized.

In addition, "○ taste" is inputted by voice, and FIG.
As shown in (B), in a state where the location of “OO taste” is displayed on the display 14 together with the surrounding map, the following selection dictionary is set. That is, FIG.
Since the map shown in (B) is displayed according to the user's request, the next instruction is to change the displayed map scale,
It is assumed that contents such as "display position scrolling", "destination", "passing point" are assumed, and it is considered that another list such as a golf course or a change in volume of sound is instructed. Hateful. Therefore, the recognition control unit 35 in this case sets the dictionary to the dictionary management unit 34 based on the state that the map is displayed in accordance with the user's request, so that the dictionary management unit 34 sets the selected dictionary limited to only the command for performing the map operation. Specify.

As described above, as the state is changed based on the operation of the user, a selection dictionary including only necessary recognition vocabulary in the changed state is set as needed. In other words, the recognized vocabulary considered necessary from the state of the navigation system 1 at that time remains valid, and the other recognized vocabulary becomes invalid as dictionary data. Speech recognition for a power vocabulary can be appropriately realized.

[Second Embodiment] In the above-described first embodiment, the base dictionary shown in FIG. 5 is used to extract only recognized vocabulary words whose genre is “meal”, thereby obtaining FIG.
In this case, the recognition vocabulary not included in the selection dictionary in FIG. 6 is completely excluded from recognition. Of course, as described above, there is no problem even if the recognition vocabulary is not a target as long as the recognition vocabulary cannot be input in the state of the system at that time. However, it cannot be said that there is no possibility of being input, but there is a recognized vocabulary having a low possibility of being input. If the recognized vocabulary is also included in the selected dictionary, the number of vocabularies that may be erroneously recognized in the dictionary data will eventually increase.

In the second embodiment in consideration of this point, the following dictionary is constructed. That is, as shown in FIG. 7, an item “priority” is further set for each recognized vocabulary. Then, in the state of the base dictionary shown in FIG. 7, all the priorities are set to 0, and in the case of the selection dictionary shown in FIG. 8, the priority is set to 10 or 5 only for a predetermined recognition vocabulary. It is. In the case shown in FIG. 8, the priority is set to 0 to 10 or 5 for the recognition vocabulary of "meal" in the genre item, so that the recognition vocabulary is more easily recognized than the other recognition vocabulary. However, in the present embodiment, among the recognized vocabulary words that are “meal” in the genre item, two recognized vocabulary words that are simply “classified title” in the detailed item, that is, “Japanese food” and “Chinese” have priority. The degree is set to 5 and the other four recognition vocabularies are set to the priority of 10.

By doing so, for example, the above-mentioned "XX"
Even though there are three recognized vocabulary words that have a high possibility of misrecognition, such as “Taste”, “Izumi-cho”, and “Izumi City”, “OO taste” has a high priority of 10 and other “Izu” Mimachi and Izumi-shi have a low priority of 0, so that the recognition vocabulary with low priority is less likely to be erroneously recognized, and as a result, erroneous recognition can be minimized. it can. Also, a recognition vocabulary set to a low priority has a low priority, but finally exists as a recognition target vocabulary in the selected dictionary, and thus can be obtained as a recognition result. Therefore, for example, the speech recognition device 30 outputs the recognition results in the order of the recognized vocabulary words having the highest priority. However, if the user instructs to output another recognition result because the recognition result is not the desired one, eventually, The recognition vocabulary having the lower priority is output as a recognition result, and may be the one desired by the user. In other words, in a situation where it is impossible to cope with the situation by completely using the option of using or not using, it is effective to set the priority in this way.

Third Embodiment The above two embodiments have been described as examples in which the voice recognition device 30 is used for inputting a user's instruction in the navigation system 1.
In addition, the present invention can be used as a voice recognition device for inputting an instruction such as a desired product in an automatic vending system.

For example, consider a cigarette vending system. In the base dictionary in this case, as shown in FIG. 9A, three items of brand, classification, and setting are set for each ID (number) 1, 2, 3,.... Here, all brands of tobacco that are theoretically available for sale by the vending machine are set as brands. That is, regardless of whether the vending machine is actually sold by the vending machine, brands of all tobacco in circulation are set. However, in reality, there is a limit to the types of cigarettes that can be sold by one vending machine, so only the brands actually sold by the vending machine are recognized.
A selection dictionary as shown in (B) is set. That is, in the item of “setting”, only brands that can actually be sold are set to “ON”, and other items are set to “OFF”. Specifically, a selection dictionary in which only the brands whose setting is “ON” is set as a recognition vocabulary is constructed, and speech recognition based on the selected dictionary is executed. This ON / O
Specifically, the setting of the FF is manually set by an administrator of the vending apparatus.

In this way, the dictionary data can be adjusted according to what brand of cigarette is selected and subject to vending, so that cigarettes of brands not handled by the vending machine are erroneously recognized. It will not be a result. Therefore, the convenience for the user of the vending machine is improved.

As in the case of the second embodiment described above, for brands not handled by the vending machine,
It is also possible to adopt a method of setting a low priority. However, in this case, brands not handled cannot be sold finally even if recognized, so it can be said that invalidating the dictionary data is more preferable than lowering the priority.

[Others] In the above-described embodiment, the speech recognition apparatus 30 has been described as an example in which the speech recognition apparatus 30 is applied to the navigation system 1 or the vending system. However, the application destination may be variously applied. In short, if the state of the system changes based on an instruction from the user, and if there is a situation where the contents of the instruction from the user are limited in a certain state, a selection dictionary that extracts the recognized vocabulary is set. As a result, an effect of minimizing erroneous recognition in that state can be obtained.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a navigation system as an embodiment.

FIG. 2 is a block diagram illustrating a schematic configuration of a speech recognition device according to the embodiment.

FIG. 3 is a flowchart illustrating a process performed by the voice recognition device according to the embodiment;

FIG. 4 is an explanatory diagram showing a specific example of a system state.

FIG. 5 is an explanatory diagram of a base dictionary set by a dictionary management unit in the voice recognition unit.

FIG. 6 is an explanatory diagram of a selected dictionary set by a dictionary management unit in the voice recognition unit.

FIG. 7 is an explanatory diagram of a base dictionary in the case of the second embodiment.

FIG. 8 is an explanatory diagram of a selection dictionary in the case of the second embodiment.

FIG. 9 is an explanatory diagram of a base dictionary and a selection dictionary in the case of the third embodiment.

[Explanation of symbols]

REFERENCE SIGNS LIST 1 navigation system 2 mechanism switch 3 remote control switch 4 operation switch unit 5 operation control unit 6 current position calculation unit 7 route calculation unit 8 guidance control unit 9 map data search unit 11 audio output control unit DESCRIPTION OF SYMBOLS 12 ... Speaker 13 ... Display control part 14 ... Display 30 ... Voice recognition device 31 ... Microphone 32 ... Input signal processing part 33 ... Recognition processing part 34 ... Dictionary management part 35 ... Recognition control part

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (Reference) G08G 1/0969 F term (Reference) 2F029 AA02 AB07 AB13 AC02 AC04 AC09 AC18 3E044 AA01 AA20 CA01 CA10 CB01 CC10 DA10 EA20 EB01 EB08 5D015 HH13 KK01 LL10 5H180 AA01 BB04 BB13 CC12 CC27 EE01 FF04 FF05 FF22 FF25 FF27 FF32 FF33

Claims (8)

[Claims] 1. A voice input means for inputting voice, and a voice input through the voice input means is compared with a plurality of comparison target pattern candidates stored in a dictionary means in advance to determine a matching degree. A voice recognition device comprising: a recognition unit that sets a high recognition result as a recognition result. Dictionary management for temporarily extracting only predetermined ones of the plurality of stored comparison target pattern candidates and validating only the extracted comparison target pattern candidates as dictionary data for recognition by the recognition means. Means for speech recognition. 2. The speech recognition apparatus according to claim 1, wherein the dictionary management unit stores the comparison target pattern, which is considered impossible as an instruction from a user in the state of the input external device. A speech recognition device, characterized by invalidating dictionary data for recognition and validating other data. 3. A voice input means for inputting voice, and a voice input via the voice input means is compared with a plurality of comparison target pattern candidates stored in advance in the dictionary means to determine a matching degree. A voice recognition device comprising: a recognition unit that sets a high recognition result as a recognition result; anda state input to an external device that handles the recognition result as user instruction content. A dictionary in which priorities are temporarily set for the stored plurality of comparison target pattern candidates, and the set comparison target pattern candidates having a higher priority are more easily used as dictionary data in recognition by the recognition means. A speech recognition device comprising management means. 4. The speech recognition apparatus according to claim 3, wherein the dictionary management unit recognizes the comparison target pattern which is considered to have a higher probability as an instruction from a user in the input state of the external device. A speech recognition device characterized by being easily used as dictionary data at the time of (1). 5. The voice recognition device according to claim 1, further comprising a navigation device, wherein the voice input means of the voice recognition device performs at least navigation processing by the navigation device. The user uses a voice to input an instruction of predetermined navigation processing-related data that needs to be specified, and the post-confirmation processing means sends the recognition result by the recognition means to the navigation device. A navigation system configured to output. 6. The navigation system according to claim 5, wherein the navigation device is capable of displaying a map or a menu on a display device, and the dictionary management means is configured to perform a map operation based on the content displayed on the display device. The navigation system, wherein the comparison target pattern considered to be impossible as an instruction from a user in display content is invalidated as dictionary data for recognition or the priority is lowered. 7. The voice recognition device according to claim 1, further comprising: a vending device, wherein the voice input unit of the voice recognition device is configured to at least select an item to be sold by the vending device. An automatic vending system, which is used for a user to input an instruction of predetermined automatic vending process related data which needs to be specified in performing a sales process by voice. 8. The automatic vending system according to claim 7, wherein the voice recognition device generates a comparison target pattern candidate relating to proper nouns of all items theoretically available for sale by the vending device. The dictionary management means is provided as dictionary data in the dictionary means, and the dictionary management means temporarily stores the dictionary data for comparison target pattern candidates relating to proper nouns other than articles that can be actually sold at the time of the vending apparatus. Vending, or lowering the priority, vending system.