JP2009175233A - Speech recognition device, navigation device, and destination setting program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech recognition device preventing incorrect recognition of input speech. <P>SOLUTION: In the speech recognition device recognizing a place indicated by the input speech by comparing the input speech with recognition grammar having information on one or more places as a speech recognition object, the speech recognition device includes a visit history list in which information on already visited places is registered, creates the recognition grammar including the already visited speech recognition objects in which vocabulary means that it is already visited and information on the already visited places registered in the visit history list are connected, and compares the input speech with the speech recognition objects included in the created recognition grammar to recognize the place indicated by the input speech. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a navigation device that guides a route to a destination, and more particularly to a navigation device that sets a destination by voice.

  Since the current car navigation system is equipped with a large-capacity hard disk drive (HDD), there are tens of millions of points that can be set as destinations. Also, in the mobile phone pedestrian navigation system, since the server manages the location information, the number of locations that can be set as destinations is tens of millions, as in the car navigation system.

  As a method of searching for a point set as a destination from information on points that can be set as a destination by a user, a telephone number of a point set as a destination, an address of a point set as a destination, or a point set as a destination There is a method of searching using a facility name (may be a part of the facility name).

  In addition, as a method for a user to input search key information such as a telephone number that is a key for searching for a point set as a destination, an input method using a keyboard displayed on a screen of a navigation system, and an input method using voice recognition It has been known.

  When the input method of the search key information is speech recognition, the navigation system includes a speech recognition unit that recognizes speech uttered by the user. The speech recognition unit learns an acoustic model of a unit standard pattern that constitutes an utterance from a large amount of speech data. The speech recognition unit generates a vocabulary group to be recognized by connecting the acoustic models of the learned unit standard patterns.

  As a unit standard pattern constituting an utterance, a syllable, a vowel stationary part, a consonant stationary part, a phoneme segment, and the like are used. Note that a phoneme segment includes transition states of a syllable, a vowel stationary part, and a consonant stationary part.

  HMM (Hidden Markov Model) technology is known as a method for expressing an acoustic model of a unit standard pattern (see Non-Patent Document 1).

  Moreover, SRGS (Speech Recognition Grammar Specification) etc. are known as a method of expressing the recognition target vocabulary group which connected the acoustic model of the unit standard pattern (refer nonpatent literature 2).

  Current speech recognition is realized by pattern matching between an acoustic model of a stochastic unit standard pattern expressed by HMM technology or the like and an acoustic model of speech input by a user.

  For this reason, the voice recognition unit may erroneously recognize the voice input by the user, and a destination different from the destination intended by the user may be selected.

  In order to suppress misrecognition of the speech recognition unit, it is conceivable to limit the vocabulary group to be recognized.

  The user designates search key information desired to be used for the search, and the voice recognition unit selects a voice recognition mode corresponding to the designated search key information. Then, the user inputs voice after designating the search key information. Note that the vocabulary group to be recognized in the selected speech recognition mode is only the vocabulary group corresponding to the designated search key information, and the vocabulary group not corresponding to the designated search key information is deleted.

  Specifically, the navigation system has a telephone number recognition mode, an address recognition mode, a facility name recognition mode, and the like as voice recognition modes. For example, the vocabulary group to be recognized in the telephone number recognition mode includes only the vocabulary group corresponding to the telephone number and does not include the vocabulary group corresponding to the address and the facility name.

  Furthermore, in the address recognition mode, the vocabulary groups to be recognized may be classified by prefecture and municipality, and in the facility name recognition mode, the vocabulary groups to be recognized may be classified by the genre of the facility.

  In this case, after specifying the search key information to be used, the user further specifies the classification. Thereby, the voice recognition unit can further reduce the vocabulary group to be recognized.

  In addition, there is known a navigation system in which a command name instructed to the navigation system is used as a speech recognition target instead of search key information of a point set as a destination (see Patent Document 1 and Patent Document 2).

  Specifically, the navigation system recognizes a voice indicating the input command name, and switches the operation of the navigation system based on the recognized command.

  There is also known a navigation system that sets a destination by using history information of a point actually visited by a user and a point history set as a destination (see Patent Literature 3 and Patent Literature 4).

  Furthermore, as a speech recognition mode, a navigation system having a destination history mode in which a vocabulary group for setting a destination includes only a vocabulary group corresponding to a point set as the destination is also known.

The navigation system switches the voice recognition mode to the destination history mode based on a command for instructing switching of the voice recognition mode. The navigation system sets the destination by comparing the voice input by the user with a vocabulary group consisting only of vocabulary groups corresponding to the point set as the destination.
JP 2004-309353 A JP 2000-259184 A JP 2006-349427 A JP 2006-300552 JP Kai-Fu Lee, "Automatic Speech Recognition: The Development of the Sphinx Recognition System", 1st edition, Kluwer Academic Pub, November 1988, p17-43 Andrew Hunt, 1 other, "Speech Recognition Grammar Specification Version 1.0", [online], March 16, 2004, W3C, [searched September 18, 2007], Internet, <URL: http: // www.w3.org/TR/speech-grammar/ ＞

  In the conventional navigation system, the user switches the voice recognition mode including the telephone number recognition mode, the address recognition mode, and the facility name recognition mode to the voice recognition mode to be used in order to reduce the vocabulary group to be recognized. After that, the voice for setting the destination was input.

  For this reason, the user has to switch the voice recognition mode before inputting the voice for setting the destination, and the switching operation is troublesome.

  An object of the present invention is to provide a navigation system that does not need to switch voice recognition modes in advance and that causes less recognition errors.

  In addition, in the conventional technology, when the name of a certain point is registered as an unvisited point and registered as a visited point, there are a large number even if the name of a visited point is spoken There was a high possibility of being misrecognized as an unvisited point.

  According to a representative aspect of the present invention, the speech that recognizes the point indicated by the input speech by comparing the input speech with a recognition grammar that uses information on one or more points as a speech recognition target. In the recognition device, the voice recognition device has a visit history list in which information on visited sites that have been visited before is registered, and is registered in the visit history list with a vocabulary meaning that the user has visited. Generating a recognition grammar including a visited speech recognition target obtained by linking the information on the visited points, and comparing the input speech with a speech recognition target included in the generated recognition grammar. It is characterized by recognizing a point indicated by the received voice.

  According to one aspect of the present invention, a recognition grammar including a visited speech recognition target is created by connecting a vocabulary meaning visited and information on a visited point registered in the visit history list. Misrecognition of recorded voice can be prevented.

<First Embodiment>
A first embodiment of the present invention will be described with reference to FIGS.

  FIG. 1 is a block diagram showing the configuration of the destination input device 100 according to the first embodiment of the present invention.

  The destination input device 100 is a device that guides the user to the destination by displaying the route from the current position to the destination input by the user on the display or by voice.

  The destination input device 100 includes a system control unit 199. The system control unit 199 includes a GPS receiver 101, an autonomous navigation sensor 102, a display 103, a microphone 104, an A / D converter 105, an HDD 106, a map information reading unit 107, a route search unit 108, a voice recognition unit 109, and a recognition grammar creation. Unit 110, point information storage unit 111, visit history management unit 112, visit history storage unit 113, and destination storage unit 114 are connected.

  The system control unit 199 includes a CPU, a ROM, a RAM, and a data bus (not shown).

  The CPU executes arithmetic processing of various programs loaded in the RAM. Various programs and various data stored in the ROM and the HDD 106 are loaded into the RAM by the CPU.

  The RAM is a work area for data used by the CPU. The CPU, ROM, and RAM are connected by a data bus.

  The GPS receiver 101 is a device that receives signals from a plurality of GPS (Global Positioning System) satellites (not shown).

  The autonomous navigation sensor 102 estimates the current position of the vehicle and the direction of the vehicle using the gyro information and the vehicle speed pulse of the vehicle on which the destination input device 100 is mounted.

  The display 103 displays map information around the current position of the vehicle and a route to a point set as a destination.

  The microphone 104 is a device that converts voice uttered by the user into an electrical signal (analog signal).

  The A / D converter 105 is a device that converts the analog signal converted by the microphone 104 into a digital signal.

  The HDD 106 stores map information, a spot information list 300, and the like. The point information includes the facility name, reading method, and genre of the point set as the destination. The point information list 300 will be described in detail with reference to FIG.

  The map information reading unit 107 reads map information stored in the HDD 106 to a RAM provided in the system control unit 199.

  The route search unit 108 calculates a movement route from the current position of the vehicle to the point set as the destination. For example, the Dijkstra method is used as a method for calculating the movement route.

  The speech recognition unit 109 compares the speech uttered by the user with the recognition vocabulary group created by the recognition grammar creation unit 110 to recognize the speech uttered by the user. The voice recognition method will be described in detail with reference to FIG.

  The recognition grammar creation unit 110 creates a recognition vocabulary group for comparison with speech uttered by the user, for example, using the SRGS format. A method of creating a recognized vocabulary group will be described in detail with reference to FIG.

  The point information storage unit 111 reads the point information stored in the HDD 106 into a RAM provided in the system control unit 199.

  The visit history management unit 112 determines a point where the user has actually visited as a visited point.

  The visit history storage unit 113 stores a visit history list 400 in which points determined to have been visited by the visit history management unit 112 are stored in the RAM or HDD 106 provided in the system control unit 199. The visit history list 400 will be described in detail with reference to FIG.

  The destination storage unit 114 records the point set as the destination in the RAM or the HDD 106.

  The destination input device 100 does not have to have all the configurations shown in FIG. Specifically, the GPS receiver 101 and the autonomous navigation sensor 102 that are used to specify the current position, the display 103 that displays the route to the destination to the user, and the voice uttered by the user is converted into an analog signal. The microphone 104 must be installed in a place where the user uses the destination input device 100.

  However, configurations other than the GPS receiver 101, the autonomous navigation sensor 102, the display 103, and the microphone 104 (that is, the A / D converter 105, the HDD 106, the map information reading unit 107, the route search unit 108, the voice recognition unit 109, the recognition grammar). The creation unit 110, the point information storage unit 111, the visit history management unit 112, the visit history storage unit 113, and the destination storage unit 114) are not provided in the destination input device 100, but the network is connected to the destination input device 100. The server 200 (see FIG. 2) connected via the server 203 (see FIG. 2) may be provided.

  In FIG. 2, a case where the destination input device 100 does not have all the configurations shown in FIG. 1 will be described as an example.

  FIG. 2 is a block diagram showing a configuration of a destination input system according to a modification of the first embodiment of the present invention. In addition, the same structure as the destination input device 100 of 1st Embodiment gives the same number, and abbreviate | omits description.

  The destination input system includes a destination input device 100 and a server 200. The destination input device 100 is connected to the server 200 via the network 203.

  The destination input device 100 includes a GPS receiver 101, an autonomous navigation sensor 102, a display 103, a microphone 104, an A / D converter 105, an HDD 106, a map information reading unit 107, a communication unit 115, and a system control unit 199.

  The communication unit 115 is an interface for connecting the destination input device 100 to the network 203.

  The server 200 includes a route search unit 108, a voice recognition unit 109, a recognition grammar creation unit 110, a point information storage unit 111, a visit history management unit 112, a visit history storage unit 113, a destination storage unit 114, a communication unit 202, and a system. A control unit 299 is provided.

  The communication unit 202 is an interface for connecting the server 200 to the network 203.

  The system control unit 299 has the same configuration as the system control unit 199 provided in the destination input device 100 shown in FIG.

  FIG. 3 is an explanatory diagram of the point information list 300 according to the embodiment of this invention.

  The point information list 300 includes an identifier 301, a facility name 302, a reading 303, a genre 304, a near 305, a latitude 306, and a longitude 307.

  In the identifier 301, a unique identifier of a point is registered. In the facility name 302, the name of the point is registered.

  In the reading 303, how to read the name of the point registered in the facility name 302 is registered. In the genre 304, the genre to which the point belongs is registered. For example, if the point is a ramen shop, “Restaurant” is registered in the genre 304.

  In near 305, true is registered in advance at a point where the name of the point does not matter as long as the point provides the same service. The nearby 305 is not used in the first embodiment, but is used in the fourth embodiment.

  The latitude of the point is registered in the latitude 306, and the longitude of the point is registered in the longitude 307.

  FIG. 4 is an explanatory diagram of the visit history list 400 according to the first embodiment of this invention.

  The visit history list 400 includes an identifier 401, a last visit date 402, a visit count 403, and a reading 404.

  In the identifier 401, a unique identifier of a visited point is registered. The identifier registered in the identifier 401 corresponds to the registered identifier registered in the identifier 301 of the spot information list 300 shown in FIG.

In the last visit date 402, the date on which the user last visited the point is registered.
The number of visits 403 registers the number of times the user has visited the point. In the reading 404, how to read the name of the facility at the point is registered.

  FIG. 5 is an explanatory diagram of input / output of various information centered on the recognition grammar creation unit 110 according to the first embodiment of this invention.

  First, a process until the voice uttered by the user is input to the voice recognition unit 109 will be described.

  The voice uttered by the user is converted by the microphone 104 into a voice signal that is an analog electric signal. Then, the microphone 104 outputs the converted audio signal to the A / D converter 105.

  The A / D converter 105 converts the input audio signal into a digital audio signal that is a digital signal, and outputs the converted digital audio signal to the audio recognition unit 109.

  Next, a process until a recognition grammar used for speech recognition by the speech recognition unit 109 is input to the speech recognition unit 109 will be described.

  First, the visit history management unit 112 registers a point visited by the user in the visit history storage unit 113 as a visited point.

  A method in which the visit history management unit 112 determines whether or not the user has visited a point will be described.

  As a first method, there is a method of determining a point set as a destination by the user as a visited point regardless of whether or not the user has actually visited. Specifically, the visit history management unit 112 reads a point set as the destination from the destination storage unit 114 that stores the destination set by the user, and determines the read point as a visited point.

  As a second method, there is a method in which a point set by the user as a destination and a point actually visited by the user is determined as a visited point. Specifically, the visit history management unit 112 reads a point set as the destination from the destination storage unit 114, and when the distance between the read point and the current position of the user is equal to or less than a predetermined value. The read point is determined as a visited point.

  As a third method, when the user does not move from the same point for a predetermined time, there is a method of determining a point having the smallest distance from this place as a visited point. Specifically, the visit history management unit 112 determines a point having the smallest distance from this place as a visited point when the movement amount of the user is continuously equal to or less than a predetermined value for a predetermined time. .

  The visit history storage unit 113 registers points determined as visited points in the visit history list 400.

  Also, the visit history storage unit 113 outputs the visit history list 400 to the recognition grammar creation unit 110 in response to a read request from the recognition grammar creation unit 110.

  The visit history management unit 112 can delete the visited points registered in the visit history list 400 stored in the visit history storage unit 113. Specifically, the user instructs a visited site to be deleted among visited sites registered in the visit history list 400. Then, the visit history management unit 112 deletes the visited points designated by the user from the visit history list 400.

  The recognition grammar creation unit 110 reads the visit history list 400 stored in the visit history storage unit 113. Then, the recognition grammar creation unit 110 creates a recognition grammar based on the read visit history list 400. The process in which the recognition grammar creation unit 110 creates the recognition grammar will be described in detail with reference to FIGS.

  The speech recognition unit 109 compares the input speech with the vocabulary group defined by the recognition grammar shown in FIG. 6 created by the recognition grammar creation unit 110 and recognizes the input speech.

  Specifically, the voice recognition unit 109 uses the digital signal input by the A / D converter 105 as a feature vector of the digital signal of the voice uttered by the user using, for example, an LPC cepstrum (Linear Predictive Coding Cepstrum). calculate.

  Then, the speech recognition unit 109 calculates an acoustic model sequence of a vocabulary group defined by the recognition grammar created by the recognition grammar creation unit 110.

  The speech recognition unit 109 executes pattern matching between the calculated feature vector of the input speech and the acoustic model string of the vocabulary group defined by the calculated recognition grammar. Then, the speech recognition unit 109 selects an acoustic model sequence of a vocabulary group that most closely approximates the input speech feature vector.

  Next, the speech recognition unit 109 recognizes the vocabulary group of the selected acoustic model sequence as the input speech, and outputs the speech recognition result to the system control unit 199. Specifically, the speech recognition result is a reading symbol string indicating how to read the recognized speech and a semantic structure indicating the meaning of the recognized speech.

  A process in which the voice recognition unit 109 outputs a reading symbol string will be described. The speech recognition unit 109 refers to the recognition grammar shown in FIG. 6 and acquires information indicating how to read the selected vocabulary group when selecting the vocabulary group that most closely approximates the input speech. Then, the voice recognition unit 109 outputs the acquired reading information to the system control unit 199 as a reading symbol string.

  A process in which the speech recognition unit 109 outputs a semantic structure will be described. The speech recognition unit 109 refers to the recognition grammar shown in FIG. 6 and acquires the identifier of the vocabulary group indicating the selected point when selecting the vocabulary group indicating the point from the vocabulary group closest to the input speech. . Then, the voice recognition unit 109 outputs the identifier of the vocabulary group indicating the acquired point to the system control unit 199 as a semantic structure.

  The system control unit 199 sets an identifier of a vocabulary group indicating a point included in the input semantic structure as a destination.

  The system control unit 199 can acquire the route to the destination by inquiring the route search unit 108 using the identifier of the vocabulary group indicating the point included in the input semantic structure as key information.

  Further, the system control unit 199 can acquire position information around the destination by inquiring the point information storage unit 111 using the identifier of the vocabulary group indicating the point included in the input semantic structure as key information.

  Then, the system control unit 199 outputs the acquired route information to the destination or position information around the destination to the display 103.

  FIG. 6 is an explanatory diagram of a recognition grammar according to the first embodiment of this invention.

  The recognition grammar creation unit 110 creates a recognition grammar based on the formats described in SRGS and SISR (Semantics Interpretation for Speech Recognition). Note that a portion enclosed by a rectangle is a portion that is dynamically created by the recognition grammar creation unit 110.

  A variable ($ gain) is substituted into a variable ($ target) to be subjected to speech recognition (601).

  In the variable ($ again), the reading of the vocabulary indicating that it has been visited, the reading of the visited point, the variable for registering the identifier of the visited point ($ visited), and “Eiku” are registered in advance. (602). In addition, as a way of reading a vocabulary indicating that a visit has been made, for example, “also”, “another” or “again” is registered. “Eiku” can be omitted.

  In the variable ($ visited), the reading of the visited point and the identifier of the visited point are registered (603). Only the variable ($ visited) is updated each time a new location is registered in the visit history list 400, and the variable ($ again) is not updated.

  When the visit history management unit 112 determines that a point has been visited and newly registers this point in the visit history list 400, the recognition grammar creation unit 110 newly sets the reading and identifier of this point as a variable ( $ Visited).

  The identifier of the visited point registered in the variable ($ vised) corresponds to the identifier registered in the identifier 301 of the point information list 300 and the identifier 401 of the visit history list 400.

  In addition, how to read the visited point and the identifier of the visited point are registered in the variable ($ visited) based on the format described in SRGS and SISR, for example, but the format described in SRGS and SISR. The format is not limited to the above, and any format may be used as long as it can be used for speech recognition.

  In the recognition grammar shown in FIG. 6, the variable ($ visited) includes the reading “Shinagawa Suizokukan” and ID “325644”, the reading “Odaiba” and ID “30673382”, and the reading as the visited points. “Midota-un” and ID “54040568” are registered.

  According to the recognition grammar shown in FIG. 6, the destination is set only when the user utters the vocabulary indicating that the user has visited and the vocabulary indicating the point registered in the recognition grammar as the visited point.

  For example, when the user utters “Go to Shinagawa Aquarium” again, the speech recognition unit 109 outputs a reading symbol string meaning “Maybe Shinagawa” to the system control unit 199, and Shinagawa A semantic structure including the aquarium ID “325644” is output to the system control unit 199.

  FIG. 7 is an explanatory diagram of a recognition grammar that recognizes English as a recognition target according to the first embodiment of this invention.

  The recognition grammar for recognition in English is created by the recognition grammar creation unit 110 based on the format described in SRGS and SISR, and is enclosed in a rectangle, similar to the recognition grammar for recognition in Japanese shown in FIG. The part is a dynamically created part.

  Regarding the recognition grammar targeting English, only the parts different from the recognition grammar targeting Japanese shown in FIG. 6 will be described, and the description of the same part as the recognition grammar targeting Japanese shown in FIG. 6 will be omitted. To do.

  In the variable ($ again), “Go to”, a variable ($ visited) in which the notation of the visited point is registered, and a vocabulary notation “again” indicating that it has been visited are registered in advance ( 702).

  In the variable ($ visited), for example, “Yankee Stadium”, “Golden Gate Bridge”, and “Disney World” are registered (703).

  The recognition grammar shown in FIG. 6 recognizes Japanese as a recognition target, but the variable ($ vised) contains the reading of the visited point and the identifier of the visited point. In the recognition grammar, the notation (visited in English) of the visited point is registered in the variable ($ visited).

  If the reading of visited points is not registered in the variable ($ visited), the speech recognition unit 109 uses the visited points registered in the recognition grammar from the point information list 300 stored in the point information storage unit 111. Get the reading of the entry that matches the notation. Then, the voice recognition unit 109 creates an acoustic model sequence of the visited points registered in the recognition grammar based on the acquired reading.

  The speech recognition unit 109 performs pattern matching between the input speech feature vector and the acoustic model string of the visited points created in this way.

  In the recognition grammar that recognizes English, it is possible to register how to read a visited point and an identifier of a visited point in a variable ($ vised).

  FIG. 8 is a flowchart of recognition grammar creation processing executed by the recognition grammar creation unit 110 according to the first embodiment of this invention.

  First, the recognition grammar creation unit 110 reads the template portion of the recognition grammar (S801). The template part of the recognition grammar is a part other than the part enclosed by the rectangle in the recognition grammar shown in FIG. 6, and specifically, a variable ($ target) and a variable ($ again).

  Next, the recognition grammar creation unit 110 reads the visit history list 400 stored in the visit history storage unit 113 (S802). Next, the recognition grammar creation unit 110 registers the identifiers 301 and the readings 303 of all points registered in the visit history list 400 in the recognition grammar variable ($ vised) (S803).

  In the process of S803, the case where the visited point identifier and the visited reading are registered in the variable ($ visited) has been described. However, the notation of the visited point is registered in the variable ($ vised). In this case, the recognition grammar creation unit 110 may register the facility names 302 of all the points registered in the visit history list 400 in the recognition grammar variable ($ vised).

  In the process of S803, there may be a case where no point is registered in the visit history list 400, and the recognition grammar creation unit 110 reads one of the read recognition grammars after reading the visit history list 400 in the process of S802. It is necessary to determine whether or not the above points are registered. Moreover, the execution order of the process of S801 and the process of S802 may be switched.

  According to the first embodiment, when the user inputs a voice to the destination input device 100 to set the destination, the destination is not set unless a vocabulary meaning that the user has visited is input. And when the vocabulary meaning visited has been input, since the point to be recognized is only the point that has been visited, the destination input device 100 is more than the case where the points to be recognized are all points. Can be prevented from being misrecognized.

(Second Embodiment)
A second embodiment will be described with reference to FIGS.

  In the first embodiment, the destination setting target is only visited points, but the second embodiment includes points other than the point where the destination setting target has been visited (that is, unvisited points).

  FIG. 9 is an explanatory diagram of input / output of various information centered on the recognition grammar creation unit 110 according to the second embodiment of this invention.

  In addition, description is abbreviate | omitted about the same structure as 1st Embodiment, and only a structure different from 1st Embodiment is demonstrated.

  The second embodiment is different from the first embodiment in that the recognition grammar creation unit 110 creates a recognition grammar with reference to the spot information list 300 stored in the spot information storage unit 111.

  Specifically, as shown in FIG. 10, the recognition grammar creation unit 110 includes visited destinations including only visited points and points visited from all points registered in the point information list 300. Create a recognition grammar that includes an unvisited destination setting target that has been deleted.

  FIG. 10 is an explanatory diagram of a recognition grammar according to the second embodiment of this invention.

  A portion enclosed by a rectangle is a portion that is dynamically created by the recognition grammar creation unit 110.

  A variable ($ again) indicating a visited destination setting target or a variable ($ first) indicating an unvisited destination setting target is substituted into a variable ($ target) to be subjected to speech recognition (1001).

  The variable ($ again) is a variable ($ visited) (1003) in which the vocabulary reading indicating that the user has visited, the reading of the visited point, and the identifier of the visited point are registered. Is registered in advance (1002). Since the variable ($ again) and the variable ($ visited) are the same as the recognition grammar (FIG. 6) of the first embodiment, description thereof is omitted.

  In the variable ($ first), a vocabulary reading indicating unvisited, a variable ($ novisited) in which the reading of an unvisited point is registered, and “eiku” are registered in advance (1004). For example, “for the first time” is registered as a way of reading a vocabulary indicating that the user has not visited. “Eiku” can be omitted.

  In the variable ($ novised), how to read an unvisited point is registered (1005).

  When the visit history management unit 112 determines that a point has been visited and registers a new point in the visit history list 400, the recognition grammar creation unit 110 registers in the visit history list 400 from the point information list 300. The registered point is deleted, and the reading of the point registered in the point information list 300 from which the point registered in the visit history list 400 has been deleted is newly registered in a variable ($ novised).

  Although not shown in FIG. 10, an identifier of an unvisited point may be registered in the variable ($ novised) as in the variable ($ visited). Note that in general, the number of unvisited points registered in the variable ($ visited) is smaller than the number of visited points registered in the variable ($ novised).

  In this way, when the speech recognition unit 109 recognizes that the user has spoken a vocabulary indicating that the user has visited, the speech recognition unit 109 selects visited sites that are registered in fewer numbers than unvisited sites. Used to recognize the name of the point uttered by the user. Thereby, the destination input device can prevent erroneous recognition, and the user does not need to switch the recognition mode.

  FIG. 11 is a flowchart of a recognition grammar creation process executed by the recognition grammar creation unit 110 according to the second embodiment of this invention.

  First, the recognition grammar creation unit 110 reads the template portion of the recognition grammar (S1101). The template part of the recognition grammar is a part other than the part surrounded by a rectangle in the recognition grammar shown in FIG. 10, and specifically, a variable ($ target), a variable ($ again), and a variable ($ first). .

  Next, the recognition grammar creation unit 110 reads the visit history list 400 stored in the visit history storage unit 113 (S1102). Next, the recognition grammar creation unit 110 reads the spot information list 300 stored in the spot information storage unit 111 (S1103).

  Then, the recognition grammar creation unit 110 deletes the spot registered in the visit history list 400 read in the process of S1102 from the spots registered in the spot information list 300 read in the process of S1103 (S1104).

  Then, the recognition grammar creation unit 110 registers the point reading 303 registered in the point information list 300 from which the visited points have been deleted, in the recognition grammar variable ($ novised) (S1105).

  Then, the recognition grammar creation unit 110 registers the identifier 301 and the reading 303 of all points registered in the visit history list 400 read in the processing of S1102 in the recognition grammar variable ($ vised) (S1106).

  Note that the recognition grammar creation unit 110 may execute the process of S1106 before executing the process of S1103.

  In addition, there are cases where points are not registered in the visit history list 400, and the recognition grammar creation unit 110 reads the visit history list 400 in the process of S1102, and then registers one or more points in the read recognition grammar. It is necessary to determine whether or not

(First Modification of Second Embodiment)
Next, a first modification of the second embodiment will be described with reference to FIG.

  In the second embodiment, an unvisited point is set as a destination setting target as a destination setting target other than the visited destination setting target. In the second embodiment, when the user sets an unvisited point as the destination, the user adds a vocabulary (“first time”) indicating that the user has not visited to the name of the unvisited point. I have to speak. For this reason, the user cannot set the destination unless the visited point and the unvisited point are clearly stored.

  In the first modification of the second embodiment, all the points including the visited points and the unvisited points are set as the destination setting targets as the destination setting targets other than the visited destination setting targets.

  FIG. 12 is an explanatory diagram of a recognition grammar according to the first modification of the second embodiment of this invention.

  A portion enclosed by a rectangle is a portion that is dynamically created by the recognition grammar creation unit 110.

  A variable ($ again) indicating a visited destination setting target or a variable ($ all) that sets all points registered in the point information list 300 as a destination setting target variable ($ target) as a target for speech recognition ($ target) Is substituted (1201).

  The variable ($ again) is a variable ($ visited) (1203) in which the reading of the vocabulary indicating that it has been visited, the reading of the visited point, and the identifier of the visited point are registered, and “eiku” Is registered in advance (1202). Since the variable ($ again) and the variable ($ visited) are the same as the recognition grammar (FIG. 6) of the first embodiment, description thereof is omitted.

  In the variable ($ all), a variable ($ allPOI) in which how to read all the points registered in the point information list 300 and “Eiku” are registered in advance (1204).

  In the variable ($ allPOI), how to read all the points registered in the point information list 300 is registered (1205). For this reason, the variable ($ allPOI) does not need to be changed even if a new location is registered in the visit history list 400, and therefore is not dynamically updated by the recognition grammar creation unit 110.

  In this way, when the speech recognition unit 109 recognizes that the user has spoken a vocabulary indicating that the user has visited, the speech recognition unit 109 selects visited sites that are registered in fewer numbers than unvisited sites. Used to recognize the name of the point uttered by the user. Thereby, the destination input device can prevent erroneous recognition, and the user does not need to switch the recognition mode.

  In the first modification of the second embodiment, when an unvisited point is set as a destination, the user does not add a vocabulary indicating unvisited to the name of the unvisited point. Even if the voice is spoken, the voice recognition unit can recognize the voiced point. However, in this case, since the voice recognition unit 109 recognizes the uttered voice using the acoustic model sequence at all points, there is a higher possibility of erroneous recognition than in the second embodiment.

(Second Modification of Second Embodiment)
Next, a second modification of the second embodiment will be described with reference to FIG.

  In the second modification of the second embodiment, when a vocabulary indicating unvisited is uttered, a vocabulary indicating visited is used by using an unvisited point as a destination setting target. If a visited point is used as a destination setting target and a vocabulary indicating that the site has not been visited or visited is not spoken, all the points registered in the point information list 300 are set as destination setting targets. Used for.

  FIG. 13 is an explanatory diagram of a recognition grammar according to the second modification of the second embodiment of this invention.

  A portion enclosed by a rectangle is a portion that is dynamically created by the recognition grammar creation unit 110.

  A variable ($ again) indicating a visited destination setting target, a variable ($ first) indicating an unvisited destination setting target, or all points are set as destination setting targets in a variable ($ target) to be subjected to speech recognition. A variable ($ all) is substituted (1301).

  The variable ($ gain) is described in detail in FIG. 6, the variable ($ first) is described in detail in FIG. 10, and the variable ($ all) is described in detail in FIG. Omitted.

(Third embodiment)
Next, a third embodiment will be described with reference to FIGS.

  In the third embodiment, erroneous recognition of the voice recognition unit 109 can be prevented by classifying visited points for each time of visiting and limiting recognition targets.

  The input / output of various information centered on the recognition grammar creation unit 110 of the third embodiment is the same as that of FIG.

  The recognition grammar creation unit 110 of the third embodiment refers to the last visit date 402 of the visit history list 400 stored in the visit history storage unit 113, classifies the visited points for each visit time, and for each visit time. Create a point to be recognized.

  FIG. 14 is an explanatory diagram of a recognition grammar according to the third embodiment of this invention.

  A portion enclosed by a rectangle is a portion that is dynamically created by the recognition grammar creation unit 110.

  A variable ($ target) indicating a visited destination setting target, a variable indicating an unvisited destination setting target ($ first), and a destination setting target of a point visited last month ($ Lastmonth) or a variable ($ lastyear) indicating a destination setting target at a point visited last year is substituted (1401).

  In the variable ($ again), a variable ($ vised) (1003) in which a vocabulary reading indicating visited, a reading of a visited point, and an identifier of a visited point are registered, and “eiku” are registered. Is registered in advance (1402). Since the variable ($ again) and the variable ($ vised) (1403) are the same as the recognition grammar (FIG. 6) of the first embodiment, description thereof is omitted.

  In the variable ($ first), a vocabulary reading indicating unvisited, a variable ($ novisited) in which the reading of an unvisited point is registered, and “eiku” are registered in advance (1404). For example, “for the first time” is registered as a way of reading a vocabulary indicating that the user has not visited. “Eiku” can be omitted. Since the variable ($ novisited) (1406) has been described in detail in the recognition grammar (FIG. 10) of the second embodiment, a description thereof will be omitted.

  In the variable ($ lastmonth), a variable ($ lastmonth_yomi) in which the vocabulary indicating the last month is registered, “e”, a variable indicating the point visited last month ($ visited_lastmonth), and “eiku” are registered. (1406).

  In the variable ($ lastmonth_yomi), for example, “sengetsu” and “Ikagetsumae” are registered (1407).

  In the variable ($ visited_lastmonth), the reading and identifier of the point visited last month are registered (1408).

  Specifically, the recognition grammar creation unit 110 refers to the date registered in the last visit date 402 of the visit history list 400 and selects an entry belonging to a month one month before the current month. Then, the recognition grammar creation unit 110 acquires the identifier registered in the identifier 401 of the selected entry and the reading registered in the reading 404. Then, the recognition grammar creation unit 110 registers the acquired reading method and identifier in a variable ($ visited_lastmonth).

  In the variable ($ lastyear), a variable ($ lastyear_yomi) in which the reading of the vocabulary indicating last year is registered, “Ita”, a variable ($ visited_lastear) indicating the place visited last year, and “eiku” are registered. (1409).

  In the variable ($ lastyear_omi), for example, “Kyonen”, “Sakunen”, and “Ichinenmae” are registered (1410).

  In the variable ($ visited_lastyear), the reading and identifier of the point visited last year are registered (1411).

  Specifically, the recognition grammar creation unit 110 refers to the date registered in the last visit date 402 of the visit history list 400 in the same manner as the process of registering the reading of the point and the identifier in the variable ($ visited_lastmonth). , An entry belonging to a year one year before the current year is selected. Then, the recognition grammar creation unit 110 acquires the identifier registered in the identifier 401 of the selected entry and the reading registered in the reading 404. Then, the recognition grammar creation unit 110 registers the acquired reading method and identifier in a variable ($ visited_lastyear).

  In the recognition grammar shown in FIG. 14, the visit time is last month and last year, but the visit time may be two years ago, two months ago, three weeks ago, and ototo.

  FIG. 15 is a flowchart of a recognition grammar creation process executed by the recognition grammar creation unit 110 according to the third embodiment of this invention.

  Note that the same processing as the recognition grammar creation processing executed by the recognition grammar creation unit 110 according to the second embodiment of the present invention shown in FIG.

  After executing the processing of S1106, the recognition grammar creation unit 110 refers to the date registered in the last visit date 402 of the visit history list 400 and selects an entry belonging to each visit time. Then, the recognition grammar creation unit 110 acquires the identifier registered in the identifier 401 of the selected entry and the reading registered in the reading 404. Then, the recognition grammar creation unit 110 registers the acquired reading and identifier in variables (variable ($ visited_lastmonth and variable ($ visited_lastear) in FIG. 14)) corresponding to each visit time (S1501).

  In this embodiment, each visit time (last month and last year) is registered in the recognition grammar in advance, but each visit time is automatically registered using the visit time creation table 1600 shown in FIG. May be.

  FIG. 16 is an explanatory diagram of a visit time creation table 1600 according to the third embodiment of this invention.

  The visit time creation table 1600 includes a visit time name 1601, a determination logic 1602, and a reading 1603.

  In the visit time name 1601, a variable for classifying the visit time is registered. Registered in the determination logic 1602 is the name of a program that executes a determination process (the process of S1501 shown in FIG. 15) for classifying the points registered in the visit history list 400 into visit times.

  Specifically, a JavaScript function name, a Java (registered trademark) class name, and a method name are registered in the determination logic 1602.

  The recognition grammar creation unit 110 calls a program that executes the determination process from the name of the program registered in the determination logic 1602. Then, the recognition grammar creation unit 110 causes the read program to execute the determination process for all points registered in the visit history list 400, and determines whether the return value of the determination process is true. When the recognition grammar creation unit 110 determines that the return value of the determination process is true, the recognition grammar creation unit 110 determines that the point is a point corresponding to the visit time. On the other hand, when the recognition grammar creation unit 110 determines that the return value of the determination process is false, the recognition grammar creation unit 110 determines that the point is not a point corresponding to the visit time.

  In the reading 1603, a vocabulary reading indicating the time corresponding to the visit time is registered.

  Next, a process in which the recognition grammar creation unit 110 creates a recognition grammar using the visit time creation table 1600 will be described with reference to FIG. Note that the variable registered in the visit time name 1601 is a variable ($ timing).

  First, the variable ($ timing) is assigned to the variable ($ target) (1401).

  In the variable ($ timing), a variable ($ timing_yomi) in which a vocabulary reading indicating a time corresponding to the visit time is registered, a variable in which “to”, a reading of a point corresponding to the visit time and an identifier are registered ( $ Visited_timing) and “eiku” are registered (1406 and 1409).

  In the variable ($ timing_yomi), a reading method registered in the reading 1603 of the visit time creation table 1600 is registered.

  In the variable ($ visited_timing), the called program executes determination processing based on the name of the program registered in the determination logic 1602 of the visit time creation table 1600, and the recognition grammar creation unit 110 corresponds to the visit time. The reading and identifier of the determined point are registered.

(Fourth embodiment)
Next, a fourth embodiment of the present invention will be described with reference to FIG.

  The input / output of various types of information centering on the recognition grammar creation unit 110 of the fourth embodiment is the same as in FIG.

  The recognition grammar creation unit 110 of the fourth embodiment has visited a point where true is registered in the near 305 even if the point where the true is registered in the near 305 of the point information list 300 is a visited point. Is not registered in the variable ($ vised) in which the reading and identifier of the point is registered.

  In near 305, true is registered at a point where the name of the point does not matter as long as it is a point providing the same service. The point where the name of the point is not limited as long as it provides the same service, specifically, includes a convenience store, a bank, a gas station, and a parking lot.

  For example, even if a point of convenience store “Convenience store Sagami-Oono station square store” is registered in the visit history list 400, the user may be any convenience store nearby from the current position. It is unlikely that you will speak a vocabulary such as “Go to Sagami-Ono Station”.

  In the present embodiment, a point that does not matter the name of the point as long as it provides the same service is not registered in the variable ($ vised) in which the reading and identifier of the visited point are registered. For this reason, vocabulary that is unlikely to be spoken by the user can be excluded from recognition targets, and erroneous recognition of the speech recognition unit 109 can be prevented.

  FIG. 17 is a flowchart of a recognition grammar creation process executed by the recognition grammar creation unit 110 according to the fourth embodiment of this invention.

  Of the recognition grammar creation processing of the fourth embodiment, the same processing as that of the recognition grammar creation processing (FIG. 8) of the first embodiment is assigned the same number, and description thereof is omitted.

  After executing the processing of S802, the recognition grammar creation unit 110 reads the reading and identifier of the location where false is registered in the near 305 among the locations registered in the visit history list 400 acquired in the processing of S802. (S1701).

  Specifically, the recognition grammar creation unit 110 acquires all identifiers registered in the identifier 401 of the visit history list 400 acquired in the process of S802. Then, the recognition grammar creation unit 110 refers to the spot information list 300, and the identifier registered in the identifier 301 of the spot information list 300 matches the identifier registered in the identifier 401 of the acquired visit history list 400. To get.

  Then, the recognition grammar creation unit 110 refers to the acquired entry and determines whether or not “false” is registered in the near 305. Next, the recognition grammar creation unit 110 registers the identifier registered in the identifier 301 of the entry in which false is registered in the near 305 and the reading registered in the reading 303 in the variable ($ vised) of the recognition grammar illustrated in FIG. To do.

  As a result, the recognition grammar creation unit 110 can register, in the variable ($ vised), the reading and identifier of the point where false is registered in the near 305 among the visited points. That is, the point where true is registered in near 305 (the point where the name of the point does not matter if it is a point providing the same service) can be deleted from the destination setting target of the visited point.

  Note that there may be cases where no points are registered in the visit history list 400, and the recognition grammar creation unit 110 reads one or more points in the read recognition grammar after reading the visit history list 400 in the process of S802. It is necessary to determine whether or not Moreover, the execution order of the process of S801 and the process of S802 may be switched.

  In the fourth embodiment, the point where the name of the point does not matter if it is a point providing the same service is not registered in the recognition grammar variable ($ vised) even if the point has been visited. However, even if the user visits a spot where the name of the spot does not matter as long as the user provides the same service, the spot may not be registered in the visit history list 400.

  Specifically, the visit history management unit 112 determines whether or not false is registered in the near 305 of the point determined to have been visited. Then, the visit history management unit 112 registers only the points where false is registered in the near 305 in the visit history list 400.

(Fifth embodiment)
Next, a fifth embodiment of the present invention will be described with reference to FIGS.

  In the fifth embodiment, a recognition grammar corresponding to each scenario is set for each scenario to which the destination input device 1800 transitions.

  FIG. 18 is a block diagram showing a configuration of a destination input device 1800 according to the fifth embodiment of the present invention. In addition, about the structure same as the destination input device 100 of 1st Embodiment among the structures of the destination input device 1800, the same number is provided and description is abbreviate | omitted.

  The destination input device 1800 includes a GPS receiver 101, an autonomous navigation sensor 102, a display 103, a microphone 104, an A / D converter 105, an HDD 106, a map information reading unit 107, a route search unit 108, a voice recognition unit 109, and a recognition grammar creation. Unit 110, point information storage unit 111, visit history management unit 112, visit history storage unit 113, destination storage unit 114, state transition control unit 1801, recognition grammar storage unit 1802, scenario storage unit 1803, and recognition grammar selection unit 1804. Is provided.

  The state transition control unit 1801 changes the state of the destination input device 1800 based on the scenario shown in FIG.

  The state of the destination input device 1800 is the state of the screen displayed on the display 103 provided in the destination input device 1800. Note that the operations that can be performed by the user and the recognition grammar used by the speech recognition unit 109 for speech recognition differ depending on the state of the transition screen.

  The recognition grammar storage unit 1802 stores the recognition grammar created by the recognition grammar creation unit 110.

  The scenario storage unit 1803 stores a scenario for changing the state of the destination input device 1800. Details of the scenario stored in the scenario storage unit 1803 will be described with reference to FIG.

  The recognition grammar selection unit 1804 selects a recognition grammar corresponding to the current state of the destination input device 1800.

  FIG. 19 is an explanatory diagram of input / output of various types of information centering on the recognition grammar creation unit 110 according to the fifth embodiment of this invention. Of the various information inputs / outputs centered on the recognition grammar creation unit 110 of the fifth embodiment, the same parts as the various information input / outputs centered on the recognition grammar creation unit 110 of the second embodiment will be described. Omitted.

  The state transition control unit 1801 changes the state of the destination input device 1800 based on the scenario stored in the scenario storage unit 1803. The state transition control unit 1801 notifies the recognition grammar selection unit 1804 of the transitioned state of the destination input device 1800.

  On the other hand, the recognition grammar creation unit 110 stores the created recognition grammar in the recognition grammar storage unit 1802.

  When notified from the state transition control unit 1801, the recognized grammar selection unit 1804 uses the recognized grammar stored in the recognized grammar storage unit 1802 to notify the destination input device 1800. Select the recognition grammar corresponding to the state of. Then, the recognition grammar selection unit 1804 notifies the speech recognition unit 109 of the selected recognition grammar.

  When the voice recognition unit 109 receives a voice signal converted into a digital signal by the A / D converter 105, the voice recognition unit 109 performs voice recognition using the recognition grammar notified by the recognition grammar selection unit 1804.

  FIG. 20 is an explanatory diagram of a scenario for changing the state of the destination input device 1800 according to the fifth embodiment of this invention.

  In the scenario shown in FIG. 20, three states in which the destination input device 1800 can make a transition are described. The three states are a top menu (2001), a destination setting state (2002), and a destination history state (2003).

  First, the top menu will be described.

  When the destination input device 1800 is in the top menu, the recognition grammar selection unit 1804 selects a recognition grammar (command) and a recognition grammar (visited) as variables to be recognized as speech recognition (2004 and 2005).

  In the recognition grammar (command), the reading of each state in which the state of the destination input device 1800 transitions from the top menu is registered. In the present embodiment, registered in the recognition grammar (command) are “Mokukiki Seiseru”, which is a way of reading the destination setting state, and “Mokukikireiki”, which is a way of reading the destination history state.

  In addition, the recognition grammar (visited) includes, as in the variable ($ again) shown in FIG. 6 of the first embodiment, how to read the vocabulary indicating visited, how to read the visited point, and the visited point. A variable ($ vised) in which an identifier is registered, and “eiku” are registered in advance.

  Also, in the variable ($ vised), the reading of the visited point and the identifier of the visited point are registered.

  Therefore, when the state of the destination input device 1800 is the top menu, the speech recognition unit 109 only recognizes a command for designating a transition destination and a visited point.

  Further, when the voice recognition unit 109 recognizes that the input voice is a command for designating a transition destination (that is, “Mokukikiseri” or “Mokukikureiki”), the destination input device 1800 Changes the state of the destination input device 1800 to a destination setting state (ID: “Search Destination”) or a destination history state (ID: “Visit Destination”) (2006 and 2007).

  Next, the destination setting state will be described.

  When the destination input device 1800 is in the destination setting state, the recognition grammar selection unit 1804 selects a recognition grammar (visited), a recognition grammar (novisited), and a recognition grammar (command_subset) (2008 to 2010).

  Since the recognition grammar (visited) is the same as the recognition grammar (visited) of the top menu, description thereof is omitted.

  Similar to the variable ($ first) shown in FIG. 10 of the second embodiment, the recognition grammar (novisited) is a variable ($ novised) for registering how to read a vocabulary indicating unvisited and how to read an unvisited point. ) And “Eiku” are registered in advance. In the variable ($ novised), how to read an unvisited point is registered.

  In the recognition grammar (command_subset), how to read a command specifying that the state of the destination input device 1800 is returned to the top menu is registered. Note that, for example, “return”, “return to top menu”, and the like are registered as how to read a command for specifying that the state of the destination input device 1800 is returned to the top menu.

  Therefore, when the destination input device 1800 is in the destination setting state, the target of speech recognition is a visited point and an unvisited point (that is, all points registered in the point information list 300), and This command returns to the top menu.

  Next, the destination history state will be described.

  When the destination input device 1800 is in the destination history state, the recognition grammar selection unit 1804 selects a recognition grammar (visited) and a recognition grammar (command_subset) (2011 and 2012).

  Since the recognition grammar (visited) is the same as the recognition grammar (visited) of the top menu, description thereof is omitted. Since the recognition grammar (command_subset) is the same as the recognition grammar (command_subset) in the destination setting state, the description is omitted.

  In the present embodiment, the state transition of the destination input device 1800 is a transition of the screen displayed on the display 103, but it is not necessarily a transition of the screen. For example, the state transition of the destination input device 1800 may be a transition of the internal state of the destination input device 1800. For example, the internal state of the destination input device 1800 changes between when the vehicle on which the destination input device 1800 is mounted is traveling and when it is stopped.

  In addition, as a trigger for the state of the destination input device 1800 to change, the user operates the screen displayed on the display 103, the user operates the remote control for the destination input device 1800, and the user There are cases where voice is input.

  According to the present embodiment, the destination input device 1800 can select a recognition grammar corresponding to each state, thereby preventing a malfunction that is not intended by the user.

  For example, when a vehicle equipped with the destination input device 1800 is running, if only the recognition grammar (visited) is set and the recognition grammar (novisited) is not set, the visited point is Do not set the destination only. As a result, the user can set only visited points in the user's memory as destinations while the vehicle is running, and can not set unvisited points as destinations by looking at the guidebook and map etc. The safety of the user can be improved.

(Sixth embodiment)
A sixth embodiment of the present invention will be described with reference to FIGS.

  The sixth embodiment relates to a destination input system development device 2100 for developing the destination input system of the first to fifth embodiments.

  FIG. 21 is a block diagram illustrating a configuration of a destination input system development device 2100 according to the sixth embodiment of this invention.

  The destination input system development device 2100 includes a system control unit 2101, a pseudo recognition unit 2102, a keyboard 2103, a display 2104, a microphone 104, an A / D converter 105, an HDD 106, a state transition control unit 1801, a voice recognition unit 109, and a recognition grammar creation. Unit 110, point information storage unit 111, visit history editing unit 2105, visit history storage unit 113, recognition grammar selection unit 1804, recognition grammar storage unit 1802, scenario storage unit 1803, and scenario editing unit 2106. In addition, the same structure as the destination input device of 1st Embodiment and 5th Embodiment gives the same number, and abbreviate | omits description.

  The system control unit 2101 includes a CPU, a ROM, a RAM, and a data bus (not shown).

  The pseudo recognition unit 2102 recognizes information input by the keyboard 2103 by comparing the information input by the keyboard 2103 with the recognition grammar selected by the recognition grammar selection unit 2116.

  The display 2104 displays an editing screen shown in FIG. The visit history editing unit 2105 edits a point designated by the user as a visited point. The visit history management unit 112 shown in FIG. 1 manages the visit history list 400 in conjunction with the destination set by the end user and the actual movement, whereas the visit history editing unit 2105 A user such as a person directly edits the visit history list 400.

  The scenario editing unit 2106 reads the scenario stored in the scenario storage unit 1803, causes the user to edit the read scenario, and stores the edited scenario in the scenario storage unit 1803.

  FIG. 22 is an explanatory diagram of input / output of various information centered on the recognition grammar creation unit 110 according to the sixth embodiment of this invention. In FIG. 22, the same reference numerals are assigned to the same portions as the input / output of various information shown in FIG. 19 of the fifth embodiment, and the description thereof is omitted.

  The pseudo recognition unit 2102 compares information input by the user via the keyboard 2103 with the recognition grammar notified from the recognition grammar selection unit 1804, and recognizes the input information. Then, the pseudo recognition unit 2102 outputs the recognized symbol string and the semantic structure of the recognized information to the system control unit 2101.

  The scenario editing unit 2106 can edit the scenario stored in the scenario storage unit 1803.

  The visit history editing unit 2105 reads the spot information list 300 stored in the spot information storage unit 111, refers to the read spot information list 300, and edits the visit history list 400 stored in the visit history storage unit 113. Is possible.

  FIG. 23 is an explanatory diagram of an edit screen 2300 according to the sixth embodiment of this invention.

  The edit screen 2300 includes a scenario editor 2301, a visit history editor 2302, a user utterance input 2303, and a recognition grammar list 2304.

  The scenario editor 2301 is a display unit for the user to edit a scenario stored in the scenario storage unit 1803. Specifically, the scenario editor 2301 displays the state transition of the destination input device and the recognition grammar set for each state.

  For example, in the scenario editor 2301 shown in FIG. 23, the transition from the “Top Menu” state to the “Search Destination” state or the “Visit Destination” state is displayed.

  In the “TopMenu” state, the recognition grammar “command” and the recognition grammar “visited” are set, and in the “SearchDestination” state, the recognition grammar “command_subset”, the recognition grammar “visited”, and the recognition grammar “novised” are set. The recognition grammar “command_subset” and the recognition grammar “visited” are set in the “Visitated Destination” state.

  In addition, by displaying the state transition of the destination input device on the scenario editor 2301, the destination input system development device 2100 can also realize a simulation when the state of the destination input device transitions.

  For example, when the destination input device is in the “TopMenu” state, the speech recognition unit 109 or the pseudo recognition unit 2102 of the destination input system development device 2100 converts the input speech or text information into a recognition grammar “command”. Compare with When the speech recognition unit 109 or the pseudo recognition unit 2102 determines that the input speech or text information matches the “destination history” or “destination setting”, the “destination history” or “destination” A value indicating “setting” is output to the state transition control unit 1801 as a semantic structure.

  Then, the state transition control unit 1801 determines a transition destination according to the value output as the semantic structure, and executes speech recognition or pseudo recognition using the recognition grammar set in the determined state. Specifically, when a value indicating “destination history” is output, the state transition control unit 1801 determines the transition destination as the “Search Destination” state.

  Then, the destination input system development device 2100 displays a screen corresponding to the state transition when the state of the destination input device transitions, and outputs guidance voice from a speaker (not shown), thereby further improving the destination input system. Simulation close to.

  When the user changes, adds, or deletes the recognition grammar set in each state displayed in the scenario editor 2301, the scenario editing unit 2106 changes the scenario stored in the scenario storage unit 1803. Add or delete.

  When changing or adding the recognition grammar set in each state, the user selects the recognition grammar to be changed and the recognition grammar to be added from the recognition grammars displayed in the recognition grammar list 2304.

  Thus, it is possible to prevent an error in which a recognition grammar that has not been created becomes a recognition grammar to be changed and a recognition grammar to be added.

  The visit history editor 2302 displays points registered in the visit history list 400. Specifically, the visit history editor 2302 displays the facility name, the last visit, and the number of times registered in the visit history list 400.

  In the facility name, among the identifiers registered in the identifier 301 of the point information list 300, registered in the facility name 302 included in the entry of the identifier that matches the identifier registered in the identifier 401 of the visit history list 400. The name of the point is displayed.

  In the last visit date, the date registered in the last visit date 402 of the visit history list 400 is displayed. The number of times registered in the number of visits 403 in the visit history list 400 is displayed as the number of times.

  When the user changes, adds, or deletes each entry displayed on the visit history editor 2302 using the keyboard 2103, the visit history editing unit 2105 changes the visit history list 400 stored in the visit history storage unit 113. , Add, or delete.

  In the user utterance input 2303, instead of the user uttering a voice, a character string of how to read the vocabulary that he was going to utter with the voice is input via the keyboard 2103.

  Then, the pseudo recognition unit 2102 compares the vocabulary reading character string input by the user with the vocabulary reading character string registered in the recognition grammar. Then, the pseudo recognition unit 2102 determines whether there is a character string that matches the character string of the vocabulary reading inputted by the user among the character strings of the vocabulary reading registered in the recognition grammar.

  The recognition grammar list 2304 displays recognition grammars that can be set for each state.

  The present invention is applied to a navigation application for a car navigation system, a PND (Personal Navigation Device), and a portable device (such as a cellular phone, a PDA, and a portable game machine) that set a destination using voice input by a user. Available.

It is a block diagram which shows the structure of the destination input device of 1st Embodiment of this invention. It is a block diagram which shows the structure of the destination input system of the modification of 1st Embodiment of this invention. It is explanatory drawing of the point information list of 1st Embodiment of this invention. It is explanatory drawing of the visit history list | wrist of 1st Embodiment of this invention. It is explanatory drawing of the input / output of various information centering on the recognition grammar preparation part of 1st Embodiment of this invention. It is explanatory drawing of the recognition grammar of 1st Embodiment of this invention. It is explanatory drawing of the recognition grammar which recognizes English of 1st Embodiment of this invention. It is a flowchart of the recognition grammar preparation process performed by the recognition grammar preparation part of 1st Embodiment of this invention. It is explanatory drawing of the input / output of various information centering on the recognition grammar preparation part of 2nd Embodiment of this invention. It is explanatory drawing of the recognition grammar of 2nd Embodiment of this invention. It is a flowchart of the recognition grammar preparation process performed by the recognition grammar preparation part of 2nd Embodiment of this invention. It is explanatory drawing of the recognition grammar of the 1st modification of 2nd Embodiment of this invention. It is explanatory drawing of the recognition grammar of the 2nd modification of 2nd Embodiment of this invention. It is explanatory drawing of the recognition grammar of 3rd Embodiment of this invention. It is a flowchart of the recognition grammar preparation process performed by the recognition grammar preparation part of 3rd Embodiment of this invention. It is explanatory drawing of the visit time creation table of 3rd Embodiment of this invention. It is a flowchart of the recognition grammar preparation process performed by the recognition grammar preparation part of 4th Embodiment of this invention. It is a block diagram which shows the structure of the destination input apparatus of 5th Embodiment of this invention. It is explanatory drawing of the input / output of various information centering on the recognition grammar preparation part of 5th Embodiment of this invention. It is explanatory drawing of the scenario which changes the state of the destination input device of 5th Embodiment of this invention. It is a block diagram which shows the structure of the destination input system development apparatus of 6th Embodiment of this invention. It is explanatory drawing of the input / output of various information centering on the recognition grammar preparation part of 6th Embodiment of this invention. It is explanatory drawing of the edit screen of 6th Embodiment of this invention.

Explanation of symbols

100 Destination Input Device 101 GPS Receiver 102 Autonomous Navigation Sensor 103 Display 104 Microphone 105 A / D Converter 106 HDD
107 Map information reading unit 108 Route search unit 109 Voice recognition unit 110 Recognition grammar creation unit 111 Point information storage unit 112 Visit history management unit 113 Visit history storage unit 114 Destination storage unit 115 Communication unit 199 System control unit 300 Point information list 400 Visit history list 1600 Visit time creation table 1800 Destination input device 1801 State transition control unit 1802 Recognition grammar storage unit 1803 Scenario storage unit 1804 Recognition grammar selection unit 2100 Destination input system development device 2101 System control unit 2102 Pseudo recognition unit 2103 Keyboard 2104 Display 2105 Visit history editing unit 2106 Scenario editing unit 2116 Recognition grammar selection unit 2300 Edit screen 2301 Scenario editor 2302 Visit history editor 2303 User utterance input 2304 Recognition grammar list

Claims (20)

In the speech recognition apparatus for recognizing a point indicated by the input voice by comparing the input voice and a recognition grammar for information on one or more points as a speech recognition target,
The voice recognition device
It has a visit history list with information on visited places that have been visited before,
Creating a recognition grammar that includes a visited speech recognition target that links a vocabulary meaning that it has been visited and information on visited points registered in the visit history list;
A speech recognition apparatus that recognizes a point indicated by the input speech by comparing the input speech with a speech recognition target included in the created recognition grammar. The voice recognition device
It has a point information list in which information on multiple points is registered,
From the information of the points registered in the point information list, by excluding the information of the visited points registered in the visit history list, to acquire the information of the unvisited points not yet visited,
Include the information of the acquired unvisited point in the recognition grammar as an unvisited speech recognition target,
The speech recognition apparatus according to claim 1, wherein a point indicated by the input speech is recognized by comparing the input speech with a speech recognition target included in the recognition grammar.   The speech recognition device does not connect the vocabulary meaning that the user has not visited yet and the information on the unvisited point, and includes the unvisited speech recognition target consisting of only the information on the unvisited point, and the visited speech recognition The speech recognition apparatus according to claim 2, wherein a recognition grammar including an object is created.   The speech recognition apparatus creates a recognition grammar including an unvisited speech recognition target obtained by connecting a vocabulary meaning that the user has not visited yet and information on the unvisited point, and the visited speech recognition target. The speech recognition apparatus according to claim 2. The voice recognition device
Classifying the visited points of the visited speech recognition target according to the visit time;
The speech recognition according to claim 1, wherein a recognition grammar including a visited speech recognition target is created by connecting a vocabulary that can specify the visit time and information on a visited point corresponding to the visit time. apparatus. The voice recognition device
Excluded points to be excluded from the visited speech recognition target are registered in advance,
The speech recognition according to claim 1, wherein even when the exclusion point is visited, the registered exclusion point is referred to and the visited exclusion point is not included in the visited speech recognition target. apparatus. The recognition grammar includes a plurality of speech recognition objects,
The voice recognition device
A plurality of states with different speech recognition targets used for comparison with the input speech, and a scenario for associating each state with the speech recognition target used for the comparison,
Transitioning states between the plurality of states;
Based on the scenario, select a speech recognition target corresponding to the transitioned state,
Creating a recognition grammar including the selected speech recognition target;
The speech recognition device according to claim 1, wherein a point indicated by the input speech is recognized by comparing the input speech with a speech recognition target corresponding to a state of the speech recognition device. .   The voice recognition apparatus according to claim 7, wherein the voice recognition apparatus changes the state of the voice recognition apparatus when a state of a vehicle on which the voice recognition apparatus is mounted is changed. The information of the visited points registered in the visit history list can be changed,
The speech recognition device is configured to change the visited speech recognition target based on the changed visit history list when the information on the visited points registered in the visit history list is changed. The speech recognition apparatus according to claim 1. In a navigation device that guides the route to a set point,
The navigation device includes:
It has a visit history list with information on visited places that have been visited before,
Accept voice input,
Creating a recognition grammar that includes a visited speech recognition target that links a vocabulary meaning that it has been visited and information on visited points registered in the visit history list;
Recognizing a point indicated by the input voice by comparing the input voice and a voice recognition target included in the created recognition grammar;
A navigation device for guiding a route from the current location of the navigation device to the recognized point. The navigation device includes:
It has a point information list in which information on multiple points is registered,
With reference to the visit history list, among the information of points registered in the point information list, obtain information of unvisited points that have not yet been visited,
In the recognition grammar including the visited speech recognition target, the information on the acquired unvisited point is included as an unvisited speech recognition target,
The navigation apparatus according to claim 10, wherein the point indicated by the input voice is recognized by comparing the input voice with a voice recognition target included in the created recognition grammar. The navigation device includes:
Classifying the visited points of the visited speech recognition target according to the visit time;
11. The navigation apparatus according to claim 10, wherein a recognition grammar including a visited speech recognition target is generated by connecting a vocabulary that can specify the visit time and information on a visited point corresponding to the visit time. . The navigation device includes:
Excluded points to be excluded from the visited speech recognition target are registered in advance,
11. The navigation device according to claim 10, wherein, even when the exclusion point is visited, the registered exclusion point is referred to and the visited exclusion point is not included in the visited speech recognition target. . The recognition grammar includes a plurality of speech recognition objects,
The navigation device includes:
A plurality of states with different speech recognition targets used for comparison with the input speech, a scenario for associating each state with the speech recognition target used for the comparison,
Transitioning states between the plurality of states;
Based on the scenario, select a speech recognition target corresponding to the transitioned state,
Creating a recognition grammar including the selected speech recognition target;
The navigation device according to claim 10, wherein the point indicated by the input voice is recognized by comparing the input voice and a voice recognition target corresponding to the state of the navigation device. The information of the visited points registered in the visit history list can be changed,
The navigation device is configured to change the visited speech recognition target based on the changed visit history list when the information on the visited points registered in the visit history list is changed. The navigation device according to claim 10. The speech recognition apparatus for recognizing the input speech recognizes the point indicated by the input speech by comparing the input speech with a recognition grammar for information on one or more points as a speech recognition target. A program to
The program is stored in the voice recognition device.
Accept voice input,
Register your visited history information in the visit history list,
Creating a recognition grammar that includes a visited speech recognition target that links a vocabulary that means having been visited and information on visited points registered in the visit history list;
A program for recognizing a point indicated by the input voice by comparing the input voice with a voice recognition target included in the created recognition grammar. The voice recognition device has a point information list in which information of a plurality of points is registered,
The program is stored in the voice recognition device.
Refer to the visit history list, to obtain the information of the unvisited points that have not yet visited among the information of the points registered in the point information list,
The recognition grammar including the visited speech recognition target includes the acquired information on the unvisited point as an unvisited speech recognition target,
The program according to claim 16, wherein a point indicated by the input voice is recognized by comparing the input voice with a voice recognition target included in the recognition grammar. The program is stored in the voice recognition device.
Classifying the visited point of the visited speech recognition target according to the visit time;
17. The program according to claim 16, wherein a recognition grammar including a visited speech recognition target is generated by connecting a vocabulary that can specify the visit time and information on a visited point corresponding to the visit time. The recognition grammar includes a plurality of speech recognition objects,
The speech recognition apparatus has a plurality of states in which speech recognition targets used for comparison with the input speech are different, and a scenario for associating each state with the speech recognition target used for the comparison,
The program is stored in the voice recognition device.
Transitioning states between the plurality of states;
Based on the scenario, the speech recognition target corresponding to the transitioned state is selected,
Creating a recognition grammar including the selected speech recognition target;
The program according to claim 16, wherein a point indicated by the input voice is recognized by comparing the input voice with a voice recognition target corresponding to the state of the navigation device. The information of the visited points registered in the visit history list can be changed,
The program is stored in the voice recognition device.
17. The visited speech recognition target is changed based on the changed visit history list when information on visited points registered in the visit history list is changed. Program.

Images