JP2005114964A - Method and processor for speech recognition - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce malfunction of an on-vehicle navigation device which vocally inputs the name of a place by improving its voice recognition precision. <P>SOLUTION: A speech recognition part outputs a plurality of results obtained by processing and recognizing an inputted voice and holds the results as candidates data in the decreasing order of matching probability. Before they are outputted to other processing parts as speech recognition results, matching between the held candidate data and map data is confirmed in order from higher matching and when candidate data matching the map data are found, the candidate data are outputted as a voice recognition result to other functions. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a technique for recognizing an address by voice.

  A vehicle-mounted navigation device that calculates the current position of the vehicle and displays it on a screen such as a liquid crystal display, as well as displaying on the screen the road on which the vehicle should travel toward the specified destination, and performing route guidance It has been known.

  In recent years, a technique for inputting an instruction by voice instead of inputting an instruction command or the like by a manual operation has been adopted in an in-vehicle navigation device.

  For example, there is a navigation device that can input various information such as a place name and a control command name as a basis of navigation by a user's real voice (see, for example, Patent Document 1). This navigation device includes a speech recognition decoder, and selects and inputs candidate words close to words represented by speech input via a microphone from previously prepared dictionary information.

JP 09-292255 A

  However, in the case of the navigation device disclosed in Patent Document 1, a place name consisting of candidate words determined to be close to the input speech by the speech recognition decoder, regardless of whether or not the place name exists on the map, Output as recognition result. Therefore, when a place name that does not exist on the map is output as a recognition result, it is not possible to perform subsequent processing such as route recognition.

  The present invention has been made to solve such a problem, and provides a speech recognition processing technology that can reliably output only a place name existing on a map as a recognition result when recognizing a place name inputted by voice. For the purpose.

  The present invention provides a speech recognition method for processing address data inputted by speech, and outputting only the address actually present in map data as a recognition result.

  For example, the speech recognition method of the present invention is a speech recognition method for recognizing address data input by speech in an apparatus comprising map data storage means for storing map data including address data for specifying a point on a map. A voice input receiving step for receiving voice input from the operator, and performing voice recognition processing on the voice received in the voice input receiving step, and setting one or more voice recognition results as candidate data in descending order of consistency. Candidate data generation step for storing in the candidate database, candidate data stored in the candidate database are extracted in order from the one with the highest consistency, and the address data matching the extracted candidate data is in the map data If there is a match, the extracted candidate data is output as the final recognition result. If there is no match, until all candidate data stored in the candidate database are extracted, the process of extracting the next most consistent data and determining the match is repeated. A recognition result determination step of outputting data indicating no match instead of the final recognition result when the address data in the map data does not match.

  Another speech recognition method of the present invention is a speech recognition method for recognizing address data input by voice in an apparatus comprising map data storage means for storing map data including address data for specifying a point on a map. A voice input receiving step for receiving voice input from the operator, a voice processing step for performing voice recognition processing on the voice received in the voice input receiving step, and address data in a specific range from the map data. The recognition candidate database that is extracted and registered with the extracted address data is compared with the speech recognition processing result obtained in the speech recognition processing step, and the most consistent one is compared with the final recognition result from the candidate database. A recognition result determination step for outputting as

  ADVANTAGE OF THE INVENTION According to this invention, in the apparatus which inputs a place name by voice, the speech recognition processing technique which can output only the place name which exists on a map reliably as a recognition result can be provided.

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

  FIG. 1 is a hardware configuration diagram of the navigation device 100 of the present embodiment. As shown in the figure, the navigation device 100 of this embodiment includes a CPU 101, a ROM 102, a RAM 103, a display device 104, an input / output interface device 105, a recording medium drive 106, and a GPS (Global Positioning System) reception. Machine 111, direction sensor 112, vehicle speed sensor 113, D / A converter 121, A / D converter, input device 123, audio input unit 124 such as a microphone, and audio output unit 125 such as a speaker. Is provided.

  The GPS receiver 111 receives a signal from a GPS satellite. The direction sensor 112 includes a geomagnetic sensor, an optical gyro, and the like, and acquires a signal that is a basis for detecting the traveling direction of the own vehicle by the sensor or the like. The vehicle speed sensor 113 generates a pulse corresponding to the rotational speed of the wheel.

  The voice input unit 124 includes a microphone and receives voice input. The D / A conversion unit 121 performs D / A conversion on the voice received by the voice input unit 124. The audio output unit 125 includes a speaker or the like, and outputs an audio signal that has been A / D converted by the A / D conversion unit 122. Further, the utterance switch 123 transmits a signal for accepting voice input for a predetermined period after the switch is pressed.

  The input / output interface device 105 sends signals acquired from the GPS receiver 111, the direction sensor 112, the vehicle speed sensor 113, the speech switch 123, the D / A conversion unit 121, and the A / D conversion unit 122 to the bus line 110.

  The recording medium drive 106 reads data from a recording medium such as a CD-ROM or DVD in which digital map data is recorded, which is loaded in a recording medium mounting unit (not shown). Here, in the digital map data recorded on the recording medium, address data is stored as information for specifying a position on the map, together with road data in a format capable of navigation processing such as general route guidance.

  The display device 104 displays various data processed for the purpose of being displayed on the navigation device 100, such as the current position of the vehicle, a road map of the surrounding area, route information to the destination, and guidance information.

  The input device 123 receives commands such as input of information to be displayed on the display device 104 and screen switching from the operator. The input device 123 includes switches, a touch panel using a display screen of the display device 104, a joystick, and the like. The input device 123 also includes an utterance switch 123a that accepts an input of an instruction to switch the navigation device 100 to a mode that accepts the voice input from the voice input unit 124 for a predetermined time.

  The CPU 101 controls the entire navigation device 100 by executing a program stored in the nonvolatile memory ROM 102. The RAM 106 is a volatile memory and serves as a work data area.

  The bus line 110 connects each component such as the CPU 101 of the navigation device 100 via a bus.

  Next, a functional configuration of the navigation device 100 according to the present embodiment will be described. FIG. 2 is a functional configuration diagram of the navigation device 100 of the present embodiment.

  As shown in the figure, the navigation device 100 of the present embodiment includes a control unit 201, an operation / input unit 211, a display / output unit 212, a current position calculation unit 221, a route search unit 222, and a recognition result. A confirmation unit 223, a speech recognition unit 224, a speech synthesis unit 225, map data 231, a recognition dictionary 232, and comparison candidate data 233 are provided.

  The current position calculation unit 221 uses a signal obtained from the GPS receiver 111, the direction sensor 112, the vehicle speed sensor 113, and map data obtained from a recording medium via the recording medium drive 106, and uses the current position of the host vehicle. The information necessary for navigation processing such as the direction of travel is calculated.

  The speech recognition unit 224 performs speech recognition processing on the speech signal input from the speech input unit 124, compares the result with dictionary data in the recognition dictionary 232, and matches a predetermined number of candidate data. The data is generated with a priority from the most likely ones and stored in the candidate database 233. Thereafter, the recognition result determination unit 232 is notified that the processing is completed.

  The candidate data stored in the candidate database 233 is to be compared with data stored as address data in the map database 231 as will be described later. Therefore, the candidate data is converted into a format that can be compared and stored and stored in the candidate database 233.

  In this embodiment, the address is classified into “prefecture name”, “city name”, “large name”, and “chome street address”, prompting the input for each component, and accepting the input A description will be given by taking, as an example, a configuration in which each processing is performed by the speech recognition unit 224 and is confirmed by a recognition result determination unit 223 described later. Accordingly, the candidate data is stored in the candidate database 233 for each component constituting the address.

  The method of inputting address data by voice and the recognition processing method are not limited to this, and the configuration is such that all the components of the address are received at once, stored in a memory, etc., and the recognition processing proceeds in appropriate units. Alternatively, for example, it can be determined as appropriate by receiving input separately for a part including a number such as “chome address” and another part (referred to as “prefectural city large name”) and proceeding with the recognition process.

  FIG. 3 shows an example of candidate data that is processed by the voice recognition unit 224 and stored in the candidate database 233. Here, data stored when “chome address” is input among the elements constituting the address data will be described as an example.

  As shown in the figure, the candidate database 233 includes priority order data 233a indicating the priority order, candidate data 233b, and determination unit data 233c indicating the component data of the part of the address determined so far.

  Here, the candidate data is shown as a character string, but is not limited thereto. As described above, the candidate data may be in a format that can be compared with the map data information.

  The recognition result determination unit 223 compares and collates with the data in the map database 231 in accordance with the priority order stored in the candidate database 233, and outputs a match as a result.

  Upon receiving the processing completion notification from the voice recognition unit 224, the recognition result determination unit 223 accesses the map database 231 and extracts the determination unit data 233c of the candidate database 233, and the next in the address determined by the determination unit data 232c. Extract component data of levels. Here, the component of the next level is, for example, “prefecture name”, “city name”, “city name” if “city name” is confirmed. If it is fixed to “Large Name”, it means “Chome Address”.

  Next, the candidate data 233b in the candidate database 233 is compared with the data extracted from the map database 231 in descending order of priority. When there is a match, it is output as a recognition result, and the added component is registered in the candidate database 233 as new confirmed part data 232c.

  For example, if the address is fixed up to “Large name” and the maximum of the “Chome address number” of the large character is “5 Chome address 7”, the candidate data 233b stored in FIG. When using the comparison and collation, the priority No. 1 “25th Street No. 6 No. 35” is determined as no match, and the process proceeds to the next priority candidate data. Then, when “2-6-6 address 35” having the second priority is extracted and collated, it is determined that there is a match and is output as a recognition result.

  Here, when there is a match, the result may be output to the display device 104 and / or the audio output unit 125 to be approved by the operator. In the case of non-approval, until the approval is received, the comparison and comparison are continued in the same manner from the next priority, and the result is output repeatedly. After receiving approval, the result is output as confirmed.

  If there is no match or no approval can be obtained even though the above processing is performed for all candidate data, the recognition result determination unit 223 indicates that the verification is unsuccessful. The data is output from the display device 104 and / or the audio output unit 125, and this is presented to the operator.

  The route search unit 222 performs route search processing using the current position, the destination, and the data in the map database 231 in accordance with the instruction received from the operator via the input device 123, and outputs the instructed result. To do.

  For example, when an instruction to perform route guidance processing is input via the input device 123 or the voice input unit 124, the current position data calculated by the current position calculation unit 221 and the purpose input via the input device 123 Road data connecting the current position and the destination using the location data or the destination data input through the voice input unit 124 and processed by the voice recognition unit 224 and the recognition result determination unit 223 Various data necessary for performing the route guidance process are acquired from the map database 231, the guidance route from the current location to the destination is determined using the Dijkstra method, and the result is transmitted to the display / output unit 212. .

  The voice synthesizer 225 converts the result processed by each functional unit into voice data for generation using existing voice synthesizer software in order to output the result from the display / output unit 212 as voice. Note that the converted audio data is output via the D / A converter 122 and the audio output unit 125 by the display / output unit 212 described later.

  The operation / input unit 211 receives a signal input via the input device 123 and an input of the signal converted into a digital signal by the A / D conversion unit 121 via the voice input unit 124 and via the control unit 201. Send to each functional unit.

  The display / output unit 212 outputs the result processed by each functional unit to the output device 104 and the audio output unit 125.

  The control unit 201 controls the functional operation of the entire navigation device 100. In the present embodiment, for example, when the utterance switch 123a of the input device 123 is pressed, control is performed so as to accept voice input via the voice input unit 124 for a predetermined time.

  Each of these functional units is realized by the CPU 101 executing a program stored in the ROM 102. The candidate database 232 is stored in the RAM 106. On the other hand, the map database 231 and the recognition dictionary 232 may be stored in a recording medium, or may be configured to be read by the recording medium drive 106 and temporarily stored in the RAM 106 when the recording medium is loaded. Good.

  Hereinafter, a processing procedure in the navigation apparatus 100 according to the present embodiment when a route search process from the current location to the destination is performed using the address of the destination input by voice will be described with reference to the drawings.

  FIG. 4 is an overall processing flow in the navigation device 100. As shown in the figure, when the control unit 201 receives an instruction for route search processing via the input device 123 or the like (step 401), the control unit 201 prompts the input of the destination, and recognizes the voice recognition unit 224 and the recognition result determination unit 223. Then, each component of the destination address data is determined (steps 402 to 405).

  When receiving a notification from the recognition result determination unit 223 that the destination has been determined, the current position calculation unit 231 calculates the current position (step 406) and notifies the route search unit 232 of the current position.

  Upon receiving the destination from the recognition result determination unit 223 and the notification of the current position from the current position calculation unit 231, the route search unit 222 uses the map database 231 to perform a route search process from the current location to the destination, A guidance route for displaying the result on the display device via the display / output unit 212 is calculated (step 407).

  Next, in the above process, the process of determining each component of the address data will be described below by taking the case of determining the “chome address” as an example.

  FIG. 5 is a processing flow in the case where the voice recognition unit 224 and the recognition result determination unit 223 determine the “chome address” from the input voice. The same processing is basically performed when determining other components of the address data.

  When the recognition result confirmation unit 223 confirms “large name”, the control unit 201 performs display or output for prompting the input of “chome address” (step 501). Specifically, the display / output unit 212 displays a screen for instructing the input of the “chome address” of the destination on the display device 104 or synthesizes a voice prompting the input of the destination to perform D / A The sound is output from the audio output unit 125 via the conversion unit 122.

  Here, the control unit 201 waits to receive a voice input permission instruction from the operation / input unit 211 via the utterance switch 123a. When the instruction is received, the control unit 201 controls the input / output interface device 105 to receive a voice input and waits for an input from the voice input unit 124.

  When the operation / input unit 211 receives a voice input via the voice input unit 124 and the A / D conversion unit 121, the operation / input unit 211 transmits the received voice signal to the voice recognition unit 224 (step 502).

  The speech recognition unit 224 performs speech recognition processing on the received speech signal, extracts candidate data related to the “chome address” in order from the highest consistency using the recognition dictionary 232, and stores the candidate data in the candidate database 233. . Then, the recognition result confirmation unit 223 is notified of the completion of storage (step 503).

  Receiving the processing completion notification, the recognition result determination unit 223 extracts the candidate data 233a stored in the candidate database 233 in order of priority, and performs comparison and collation with the data extracted from the map database 231 (steps 504, 505, and 507). If they match, the candidate data is output to the display device 104 or the voice output unit 125 to obtain approval (steps 505 and 506).

  When the approval is obtained, the recognition result determination unit 223 adds the candidate data as a result to the determination unit data 233c of the comparison candidate database 233. At this time, if the confirmed component is the last component constituting the address data, that is, “chome address”, the confirmed unit data 233c and the data indicating the end of processing are obtained as the route search unit 222 and the current The position calculation unit 231 is notified. In the case of other components, the control unit 201 is notified to display a screen that prompts the input of the next level component of the address data.

  Here, in the case where they do not match in step 505 and in the case where approval is not obtained in step 506, the recognition result determination unit 223 extracts candidate data 233a of the next priority, and performs comparison and collation in step 504. Repeat the process.

  If there is no match even though all candidate data have been compared and collated, or there is no data that can be approved, the data indicating “voice recognition failure” is displayed on the display device 104 or the voice output. The data is displayed and output on the unit 125, and the process is terminated.

  Here, the processing may be terminated without returning to step 501 and prompting for the input of the “chome address” again.

  As described above, the navigation apparatus 100 according to the first embodiment assigns a plurality of candidate data obtained as a result of the speech recognition processing to the address data input by speech, in order of priority from the data having high consistency at the time of recognition. However, it is configured such that the one having higher priority and matching the map data is output to the subsequent processing function as the final result. For this reason, the processing result of the address data inputted by voice always exists in the map data.

  Therefore, the navigation device 100 according to the present embodiment does not output an address that does not exist by erroneously recognizing the address input by voice, and outputs it to a subsequent processing function. There is no inconvenience in processing.

  In the first embodiment, the speech recognition unit 224 stores the speech recognition results in the candidate database 233 with priorities in descending order of consistency, and the recognition result determination unit 223 matches them according to the priorities. However, the method is not limited to this. For example, the voice recognition unit 224 stores a result of a predetermined consistency or higher in the candidate database 233 without assigning a priority order, and the recognition result determination unit 223 presents the match to the operator and obtains approval. May be configured to be the final result.

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

  The hardware configuration of the navigation device 100 of this embodiment is basically the same as that of the first embodiment.

  FIG. 6 shows a functional configuration diagram of the navigation device 100 of the present embodiment. As shown in the figure, the navigation device 100 of the present embodiment includes a control unit 301, an operation / input unit 311, a display / output unit 312, a current position calculation unit 321, a route search unit 322, and candidate data. A generation unit 323, a speech recognition unit 324, a speech synthesis unit 325, a map database 331, a conversion database 332, and a recognition candidate database 333 are provided.

  Among these function units, except for the candidate data generation unit 323, the speech recognition unit 324, and the recognition candidate database 333, basically the same functions as the function units having the same names in the first embodiment are provided. In addition, each of these functional units is realized by the CPU 101 executing a program stored in the ROM 102, and the map data 331 and the recognition dictionary 332 are recorded in a recording medium and recorded as necessary. The same is true in that it may be read out via the medium drive or read out when the recording medium is loaded and temporarily stored in the RAM 103. The recognition candidate database 333 is stored in the RAM 103.

  Hereinafter, the candidate data generation unit 323, the speech recognition unit 324, the conversion database 332, and the recognition candidate database 333 will be described.

  The candidate data generation unit 323 extracts candidate data from the map database 331. Specifically, the map database 331 is accessed, and a component at the next level of the component is extracted from an address made up of components that have already been determined and held by the voice recognition unit 324 described later. Then, the extracted data is converted into phonetic character string data indicating how to read using the conversion database 332, and the phonetic character string data is registered in the recognition candidate database 333 as candidate data.

  In addition, the component of the next level of the component is the same as in the first embodiment, for example, the address data includes “prefecture name”, “city name”, “large name”, chome address, If it has been confirmed up to “prefecture name”, “city name”, “city name”, “large name”, “large name” ”Is the“ chome street address ”.

  Here, the conversion database 332 will be described. The conversion database 332 stores a correspondence table for converting the address data stored in the map database 331 into phonetic character string data indicating how to read the address data.

  And the recognition candidate database 333 stores the candidate data which consists of the phonetic character string produced | generated in the candidate data production | generation part 323 as mentioned above.

  The voice recognition unit 334 performs voice recognition processing on the address name input by voice from the voice input unit 124, compares the result with candidate data stored in the recognition candidate database 333, and has the highest consistency. Output the result. In addition, when the address data is determined for each component, the already determined component is held.

  Here, in the present embodiment, the phrase having the highest consistency is assumed to have the most consistent phonetic characters including the arrangement order of the phonetic character strings. This can be determined as appropriate.

  In the present embodiment, the voice recognition unit 334 compares and collates phonetic character strings, but is not limited thereto. For example, the conversion database 332 stores data for converting data having a format equivalent to the address data stored in the map database into a predetermined voice pattern. The candidate data generation unit 323 generates a voice pattern of the address data in the map database 331 using the conversion database 332. The voice recognition unit 334 may perform voice recognition processing on the input voice, generate a voice pattern equivalent to the voice pattern registered in the recognition candidate database 333, and compare and collate both voice patterns. Good.

  Moreover, what is output as a result is assumed to have the highest consistency, but is not limited thereto. For example, it is possible to output a result having a predetermined consistency or more as a result, and select a result obtained from the operator as a final result.

  Next, a processing procedure in the navigation device 100 according to the present embodiment when a route search process from the current location to the destination is performed using the address of the destination input by voice will be described with reference to the drawings.

  In addition, since the flow of the whole process which receives the instruction | indication of a route search process via the input device 123 etc., determines a destination and a present location, and performs a route search is the same as that of 1st embodiment, it is here. I will not explain. Here, in the processing flow shown in FIG. 4 of the first embodiment, the process of determining each component of the address data will be described using an example of determining “chome address” as an example.

  FIG. 7 is a processing flow when the speech recognition unit 324 and the candidate data generation unit 323 determine the “chome address” from the input speech. The same processing is basically performed when determining other components of the address data.

  When the voice recognition unit 324 confirms the “large name”, the control unit 301 performs display or output for prompting the input of the “chome address” (step 701). Specifically, the display / output unit 312 displays a screen for instructing the input of the “chome address” of the destination on the display device 104 or synthesizes a voice prompting the input of the destination to perform D / A The sound is output from the audio output unit 125 via the conversion unit 122.

  Here, the control unit 301 waits to receive a voice input permission instruction from the operation / input unit 311 via the utterance switch 123a, and generates candidate data for “chome address” in the candidate data generation unit 323. Instruct.

  Upon receiving support for candidate data generation, the candidate data generation unit 323 accesses the map database 331 and extracts all large-size “chome street address” data already determined by the voice recognition unit 324 from the map database 331 (step 702). ). The extracted “chome address” data is converted into candidate data using the conversion database 332 (step 703) and stored in the recognition candidate database 333 (step 704).

  On the other hand, when the control unit 301 receives a voice input acceptance permission instruction via the speech switch 123 a, the control unit 301 controls the input / output interface device 105 to accept voice input and waits for input from the voice input unit 124.

  When the operation / input unit 311 receives a voice input via the voice input unit 124 and the A / D conversion unit 121, the operation / input unit 311 transmits the received voice signal to the voice recognition unit 324.

  The speech recognition unit 324 performs speech recognition processing on the received speech signal (step 706), compares the result with the candidate data stored in the recognition candidate database 333, and recognizes the result with the highest consistency as the recognition result. Is output to the display device 104 or the audio output unit 125 (step 707), and an input for approval is waited (step 708).

  When the approval instruction is received, the voice recognition processing unit 324 adds to the already determined component data held by itself.

  At this time, if the confirmed component is the last component constituting the address data, that is, “chome address”, the address data composed of each confirmed component data and the data indicating the end of processing, To the route search unit 222 and the current position calculation unit 231. In the case of other components, the control unit 301 is notified to display a screen that prompts input of the next level component of the address data, and the already determined components are notified to the candidate data generation unit 323.

  On the other hand, if the approval is not obtained in step 708, the voice recognition unit 324 causes the display device 104 or the voice output unit 125 to display and output data indicating “voice recognition failure”, and ends the process. Here, it may be configured to return to step 701 again without prompting the process to be prompted to input the “chome address”. Alternatively, it may be configured to return to step 706, adopt the next most consistent as the recognition result, and perform the processing after step 707.

  In the present embodiment, the address is input for each component and the approval is obtained. However, the present invention is not limited to this. A plurality of continuous components such as “prefecture name” and “city name” may be combined to perform voice recognition processing for each unit. In this case, candidate data is generated and held for each unit.

  In addition, in the case of a component limited by numbers such as chome address data, all the data of “chome address” of the corresponding “large name” existing in the map database 331 is extracted, and candidate data is generated. Instead, the maximum value of each chome, address, and number may be extracted, and candidate data that can recognize each chome, address, and number below the maximum value may be prepared.

  As described above, the navigation device 100 according to the second embodiment extracts candidate data for speech recognition from map data in advance, and confirms the recognition result therein. In speech recognition, candidate data is limited, so that the recognition rate is improved. Furthermore, since only the data in the range existing in the map data is the candidate data, the address data output as the recognition result always exists in the map data. Therefore, a result that does not exist in the map data is passed to the subsequent processing function, and malfunctions are not induced.

  In addition, in said 1st embodiment and 2nd embodiment, although demonstrated taking the vehicle-mounted navigation apparatus as an example, the navigation apparatus of the application object of this invention is not restricted to vehicle-mounted. It can also be applied to a hand carry navigation device. Further, the present invention is not limited to a navigation device, and can be applied to a general device that inputs an address when specifying a place name, such as a map display device.

FIG. 1 is a hardware configuration diagram of the navigation device according to the first embodiment. FIG. 2 is a functional configuration diagram of the navigation device according to the first embodiment. FIG. 3 shows an example of data stored in the comparison candidate database of the first embodiment. FIG. 4 is a processing flow of the navigation device of the first embodiment. FIG. 5 is a processing flow in the case of determining the chome address from the input voice in the first embodiment. FIG. 6 is a functional configuration diagram of the navigation device according to the second embodiment. FIG. 7 is a processing flow in the case of determining the chome address from the input voice in the second embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Navigation device, 101 ... CPU, 102 ... ROM, 103 ... RAM, 104 ... Display device, 105 ... Input / output interface device, 106 ... Storage medium drive, 110 ... Bus, 111 ... GPS receiver, 112 ... Direction sensor, 113 ... Vehicle speed sensor, 121 ... A / D converter, 122 ... D / A converter, 123 ...・ Input device 124 ... Audio input unit 125 ... Audio output unit 201 ... Control unit 211 ... Operation / input unit 212 ... Display / output unit 221 ... Current Position calculation unit, 222 ... Route search unit, 223 ... Recognition result determination unit, 224 ... Speech recognition unit, 225 ... Speech synthesis unit, 231 ... Map database, 232 ... Recognition dictionary 233 ... Comparison candidate database

Claims (8)

A speech recognition method for recognizing address data input by voice in an apparatus comprising map data storage means for storing map data including address data for identifying a point on a map,
A voice input receiving step for receiving voice input from an operator;
A candidate data generation step of performing speech recognition processing on the speech received in the speech input reception step, and storing one or more speech recognition results in a candidate database in order from a highly consistent one;
Candidate data stored in the candidate database is extracted in descending order of the consistency, and it is determined whether the address data matching the extracted candidate data is in the map data. If there is, the extracted candidate data is output as the final recognition result, and if there is no match, the next most consistent data is extracted until all candidate data stored in the candidate database are extracted. Recognizing the match, and if all of the candidate data does not match the address data in the map data, a recognition result confirmation step of outputting data indicating no match instead of the final recognition result;
A speech recognition method comprising: Mounted in an apparatus comprising map data storage means for storing map data including address data consisting of prefecture city large name data and chome street address data for specifying a point on the map, and among the address data, the prefecture city A speech recognition method for recognizing the chome address data of address data for which capitalized name data is confirmed,
A voice input receiving step for receiving voice input from an operator;
A candidate data generation step of performing voice recognition processing on the voice received by the voice input receiving means, and storing one or more voice recognition results in a candidate database in order from a highly consistent one;
Candidate data stored in the candidate database is extracted in order from the one with the highest consistency, and the chome street address data matching the extracted candidate data is determined by the already established prefecture city large name data. It is determined whether or not the data is in the street address data in the address. If there is a match, the extracted candidate data is output as a final recognition result. If there is no match, the candidate database is stored in the candidate database. Until all candidate data is extracted, the process of extracting the next most consistent data and determining the match is repeated, and all the candidate data are within the address determined by the already established prefecture-city large name data. If the data does not match the chome street address data, a recognition result confirmation step for outputting data indicating no match instead of the final recognition result,
A speech recognition method comprising: A speech recognition method for recognizing address data input by voice in an apparatus comprising map data storage means for storing map data including address data for specifying a point on a map,
A voice input receiving step for receiving voice input from an operator;
A voice processing step for performing voice recognition processing on the voice received in the voice input receiving step;
The address data within a specific range is extracted from the map data, the recognition candidate database in which the extracted address data is registered is compared with the speech recognition processing result obtained in the speech recognition processing step, and the most consistent A recognition result determination step for outputting a higher one as a final recognition result from the candidate database;
A speech recognition method comprising: Mounted in an apparatus comprising map data storage means for storing map data including address data consisting of prefecture city large name data and chome street address data for specifying a point on the map, and among the address data, A speech recognition method for recognizing the street address data of address data for which capitalized name data is fixed,
A voice input receiving step for receiving voice input from an operator;
A voice processing step of performing voice recognition processing on the voice received by the voice input receiving means;
Extracting the street address data in the address determined by the already determined prefecture city large name data from the map data, the recognition candidate database in which the extracted street address data is registered, and obtained in the voice processing step. A speech recognition processing result, and a recognition result confirmation step for outputting the highest consistency as a final recognition result from the candidate database;
A speech recognition method comprising: A speech recognition processing device mounted on an apparatus comprising map data storage means for storing map data including address data for specifying a point on a map,
Voice input receiving means for receiving voice input from an operator;
Candidate data generating means for performing voice recognition processing on the voice received by the voice input receiving means, and storing one or more voice recognition results in a candidate database as candidate data in descending order of consistency;
Candidate data stored in the candidate database is extracted in descending order of the consistency, and it is determined whether the address data matching the extracted candidate data is in the map data. If there is, the extracted candidate data is output as the final recognition result, and if there is no match, the next most consistent data is extracted until all candidate data stored in the candidate database are extracted. Recognizing the match, and if all of the candidate data does not match the address data in the map data, a recognition result determination unit that outputs data indicating no match instead of the final recognition result;
A speech recognition processing apparatus comprising: Mounted in a navigation device comprising map data storage means for storing map data including address data consisting of prefectural city capital name data and chome street address data for identifying points on the map, and among the address data, the prefecture A speech recognition processing device for recognizing the street address data of the address data for which the city large name data is fixed,
Voice input receiving means for receiving voice input from an operator;
Candidate data generating means for performing voice recognition processing on the voice received by the voice input receiving means, and storing one or more voice recognition results in a candidate database as candidate data in descending order of consistency;
Candidate data stored in the candidate database is extracted in order from the one with the highest consistency, and the chome street address data matching the extracted candidate data is determined by the already established prefecture city large name data. It is determined whether or not the data is in the street address data in the address. If there is a match, the extracted candidate data is output as a final recognition result. If there is no match, the candidate database is stored in the candidate database. Until all candidate data is extracted, the process of extracting the next most consistent data and determining the match is repeated, and all the candidate data are within the address determined by the already established prefecture-city large name data. If the data does not match the street address data, the recognition result confirmation means for outputting data indicating no match instead of the final recognition result,
A speech recognition processing apparatus comprising: A speech recognition processing device mounted on a navigation device comprising map data storage means for storing map data including address data for specifying a point on a map,
Voice input receiving means for receiving voice input from an operator;
Voice processing means for performing voice recognition processing on the voice received by the voice input receiving means;
Candidate database generating means for extracting address data in a specific range from the map data and generating a recognition candidate database in which the extracted address data is registered;
A recognition result determination unit that compares the result obtained in the voice processing unit with the candidate database, extracts the highest consistency as a final recognition result from the candidate database, and outputs the final recognition result;
A speech recognition processing apparatus comprising: It is mounted on a navigation device that includes map data storage means for storing map data including address data consisting of prefecture city large name data and chome street address data for specifying a point on the map, and among the address data, the prefecture A speech recognition processing device for recognizing the street address data of the address data for which the city large name data is fixed,
Voice input receiving means for receiving voice input from an operator;
Voice processing means for performing voice recognition processing on the voice received by the voice input receiving means;
Candidate database generating means for extracting chome street address data in an address determined by the already determined prefecture city large name data from the map data, and generating a recognition candidate database in which the extracted chome street address data is registered;
A recognition result determination unit that compares the result obtained in the voice processing unit with the candidate database, extracts the highest consistency as a final recognition result from the candidate database, and outputs the final recognition result;
A speech recognition processing apparatus comprising:

Images