JP2012088370A - Voice recognition system, voice recognition terminal and center - Google Patents

Abstract

PROBLEM TO BE SOLVED: To suppress unnecessary increase of the size of a recognition dictionary in a voice recognition terminal while expanding the recognition dictionary of the voice recognition terminal by using a recognition dictionary of a center.SOLUTION: Selecting a case in which a voice recognition terminal 1 failed in performing voice recognition of user's utterance voice by the own device, a voice recognition terminal 1 receives a word of the voice recognition result of the utterance voice and voice feature data for comparison of the word in the recognition dictionary of the center side from the center 2, and adds the received voice recognition feature data for comparison of the word to the recognition dictionary of the terminal side.

Description

  The present invention relates to a voice recognition system, a voice recognition terminal, and a center.

  Conventionally, there is a voice recognition system that performs voice recognition at a voice recognition terminal and a center. For example, in Patent Document 1, when a voice recognition terminal cannot recognize a user's uttered voice, voice data of the uttered voice is transmitted to the center, and the center performs voice recognition using the voice data, and the recognition is performed. A technique for transmitting a result to a voice recognition terminal is described.

  Patent Document 2 describes a technology in which a speech recognition terminal keeps its own recognition dictionary up-to-date by receiving a recognition dictionary from the center.

JP 2004-184858 A JP 2000-105681 A

  In the technique of Patent Document 1, since the dictionary of the voice recognition terminal does not change, the frequency of relying on the center does not change even after a long time. Therefore, there is a high possibility that the recognition result cannot be obtained in a situation where communication with the center is not possible.

  On the other hand, since the server recognition dictionary is often large in size, as in Patent Document 2, if the voice recognition terminal receives and uses the entire server recognition dictionary, the storage area of the voice recognition terminal is under pressure. End up. In addition, since the speech recognition terminal has a large number of recognition words, many words that are not used are included in the speech recognition terminal, which may reduce the recognition accuracy.

  The present invention has been made in view of the above points, and an object thereof is to suppress a useless increase in the size of the recognition dictionary in the speech recognition terminal while expanding the recognition dictionary of the speech recognition terminal using the recognition dictionary in the center.

  In order to achieve the above object, an invention according to claim 1 is a speech recognition system including a speech recognition terminal (1) and a center (2), wherein the speech recognition terminal (1) is a vehicle-side recognition dictionary. And a terminal-side control circuit unit (15). The center (2) includes a center-side recognition dictionary storage unit (in which a center-side recognition dictionary is recorded). 22) and a center side control circuit unit (23), wherein the center side recognition dictionary has voice feature data for comparison of words that the terminal side recognition dictionary does not have, and the terminal side control circuit unit ( 15) acquires voice feature data based on the user's uttered voice (110), and based on the comparison between the acquired voice feature data and the voice feature data for comparison of each word in the vehicle-mounted side recognition dictionary, A single equivalent of speech Terminal-side speech recognition means (110, 115) for extracting utterances, and the terminal-side speech recognition means (15a, 110, 115) inquires about speech data based on the uttered speech based on the failure to extract words. Inquiry transmission means (130) for transmitting to the center (2) as voice data, and the center side control circuit unit (23) is based on the inquiry voice data transmitted from the voice recognition terminal (1). Center-side speech recognition means (23b) for extracting a word corresponding to the uttered speech based on comparison between the speech feature data and comparison speech feature data for each word in the center-side recognition dictionary; A recognition result including the word extracted by the voice recognition means (23b) and the comparison voice feature data of the word in the center side recognition dictionary, Response means (23a) for transmitting to the voice recognition terminal (1), and the terminal-side control circuit section (15) further includes a comparison voice feature included in the recognition result received from the center (2). A speech recognition system comprising dictionary updating means (150) for additionally registering data in the terminal side recognition dictionary as comparison voice feature data for words included in the received recognition result.

  As described above, the voice recognition terminal (1) selects the case where the user's speech voice has failed to be recognized by the user's own device, and from the center (2), along with the word of the voice recognition result of the speech voice, The comparison voice feature data of the word in the recognition dictionary is received, and the received comparison voice feature data of the word is added to the terminal side recognition dictionary.

  By doing in this way, the comparison voice feature data of the words that need to be recognized by the speech recognition terminal at least once is selectively registered in the speech recognition terminal. The possibility that such a word needs to be recognized again is higher than a word that does not need to be recognized once, so the comparative voice feature data of the word is relatively likely to be wasted. Low. Therefore, it is possible to reduce the increase in the size of the recognition dictionary in the speech recognition terminal while expanding the recognition dictionary of the speech recognition terminal using the center recognition dictionary.

  The invention described in claim 2 is a voice recognition terminal that communicates with the center (2), and includes a vehicle-side recognition dictionary storage unit (14) in which a vehicle-side recognition dictionary is recorded, and a terminal-side control circuit unit ( 15), and the terminal side control circuit unit (15) acquires voice feature data based on the user's uttered voice (110), and acquires the acquired voice feature data and each word in the in-vehicle side recognition dictionary Terminal-side speech recognition means (110, 115) for extracting a word corresponding to the uttered speech based on comparison with the comparison speech feature data, and the terminal-side speech recognition means (15a, 110, 115) Inquiry transmission means (130) for transmitting voice data based on the uttered voice to the center (2) as inquiry voice data based on the failure to extract the voice, and the center (2) Based on the comparison between the voice feature data based on the inquiry voice data transmitted from the voice recognition terminal (1) and the voice feature data for comparison of each word in the center side recognition dictionary, a word corresponding to the uttered voice is extracted. The recognition result received from the center (2) when the recognition result including the extracted word and the comparison speech feature data of the word in the center-side recognition dictionary is transmitted to the speech recognition terminal. Dictionary update means (150) for additionally registering the comparison speech feature data contained in the terminal-side recognition dictionary as comparison speech feature data for the words contained in the received recognition result. Is a voice recognition terminal. As described above, the features of the invention of the speech recognition system can also be understood as the features of the invention of the speech recognition terminal.

  The invention according to claim 3 is a center that communicates with the voice recognition terminal (1), and a center side recognition dictionary storage unit (22) in which a center side recognition dictionary is recorded, and a center side control circuit unit ( 23), and when the voice recognition terminal (1) transmits voice data based on the user's uttered voice to the center (2) as inquiry voice data, Center-side speech recognition means (23b) for extracting a word corresponding to the uttered speech based on comparison between the speech feature data based on the inquiry speech data and comparison speech feature data for each word in the center-side recognition dictionary. ), The word extracted by the center side speech recognition means (23b), and comparison voice feature data of the word in the center side recognition dictionary, The speech recognition terminal (1) transmits the comparison speech feature data included in the recognition result to the speech recognition terminal (1) as comparison speech feature data for a word included in the recognition result. And a response means (23a) for additionally registering in the terminal side recognition dictionary. As described above, the features of the invention of the voice recognition system can also be understood as the features of the invention of the center.

  In addition, the code | symbol in the bracket | parenthesis in the said and the claim shows the correspondence of the term described in the claim, and the concrete thing etc. which illustrate the said term described in embodiment mentioned later. .

It is a mimetic diagram of a voice recognition system concerning an embodiment of the present invention. It is a block diagram of the vehicle equipment 1 and the center 2. FIG. 4 is a flowchart of processing executed by a terminal-side control circuit unit 15.

  Hereinafter, an embodiment of the present invention will be described. FIG. 1 schematically shows a speech recognition system according to the present embodiment. This voice recognition system includes an in-vehicle device 1 (corresponding to an example of a voice recognition terminal) mounted on a vehicle and a center 2 installed in a remote place (for example, in a building) outside the vehicle. The communication path between the in-vehicle device 1 and the center 2 may be any type. For example, communication may be enabled via a communication path such as a wireless base station or a wide area network (for example, the Internet), or direct wireless communication may be possible.

  In the present embodiment, the in-vehicle device 1 tries to recognize the user's utterance voice in the vehicle, and when the recognition fails, the in-vehicle device 1 transmits voice data based on the utterance voice to the center 2 as inquiry data. Speech recognition is performed on speech data based on the data, the word obtained as a result of speech recognition, and vocabulary data of the word in the speech recognition dictionary (character string of the word and speech feature data for comparison of the word (described later) Are transmitted to the in-vehicle device 1. The in-vehicle device 1 adds the received vocabulary data to the speech recognition dictionary of the own device as the vocabulary data of the received word.

  Hereinafter, the configuration and operation of such a speech recognition system will be described. In FIG. 2, the structure of the vehicle equipment 1 and the center 2 is shown with a block diagram. As shown in this figure, the in-vehicle device 1 includes an input device 11, an output device 12, a terminal-side communication unit 13, a terminal-side recognition dictionary storage unit 14, and a terminal-side control circuit unit 15.

  The input device 11 is a microphone that receives an input of uttered voice uttered by a user in the vehicle and outputs an audio signal of the received uttered voice to the terminal-side control circuit unit 15. The output device 12 is a device that provides information to the user, such as a display that outputs an image and a speaker that outputs sound.

  The terminal side communication unit 13 is a known wireless communication device for communicating with the center 2 and other communication devices. The terminal-side control circuit unit 15 can communicate with the center 2 and other communication devices using the terminal-side communication unit 13. Examples of other communication devices include a Web server connected to a wide area network.

  The terminal-side recognition dictionary storage unit 14 is a nonvolatile writable storage medium (for example, a magnetic storage medium or a flash memory) in which the terminal-side recognition dictionary is recorded. The terminal-side recognition dictionary has a plurality of vocabulary data each corresponding to one word, and each vocabulary data includes comparison voice feature data of the word and character string data of the word. The comparison voice feature data is data for comparison with the voice feature data based on the speech voice.

  For example, the comparison voice feature data of the vocabulary data corresponding to the word “convenience store” is data indicating typical features of sound when a person utters “convenience store” (for example, the feature amount and frequency of the sound) Spectrum data, time-series sampled intensity, etc.). In addition, the vocabulary data corresponding to the word “convenience store” includes other names of “convenience store” (“convenience store”, “convenience store” as voice feature data for comparison of the vocabulary data corresponding to the word “convenience store”. ”Etc.) may be further included as data indicating typical characteristics of the sound when the utterance is uttered (may be the characteristic amount of the sound, the frequency spectrum, or the intensity sampled in time series). . Further, the character string data of the vocabulary data corresponding to the word “convenience store” is a character string “convenience store”.

  The terminal-side control circuit unit 15 is realized by a microcomputer including a CPU, RAM, ROM, I / O, and the like, and various processes are realized by the CPU executing programs recorded in the ROM. Details of the processing will be described later.

  The center 2 includes a center side communication unit 21, a center side recognition dictionary storage unit 22, and a center side control circuit unit 23. The center side communication unit 21 is a well-known communication interface device for communicating with the in-vehicle device 1. The center side control circuit unit 23 can communicate with the in-vehicle device 1 using the center side communication unit 21.

  The center-side recognition dictionary storage unit 22 is a nonvolatile writable storage medium (for example, a magnetic storage medium or a flash memory) in which the center-side recognition dictionary is recorded. Similar to the terminal-side recognition dictionary, the center-side recognition dictionary has a plurality of vocabulary data each corresponding to one word, and each vocabulary data includes voice comparison data for the word and a character string of the word Contains data.

  However, the center-side recognition dictionary is an extended recognition dictionary that has comparison voice feature data for all words included in the terminal-side recognition dictionary and also includes a large number of comparison voice feature data for words that the terminal-side recognition dictionary does not have. is there. Therefore, the data size of the center side recognition dictionary is much larger than the data size of the terminal side recognition dictionary (for example, 100 times or more).

  If an attempt is made to download the entire center side recognition dictionary to the in-vehicle device 1, the storage capacity is compressed, and in some cases, there is no storage capacity to store the entire center side recognition dictionary. In addition, since the in-vehicle device 1 has a large number of recognition words, many words that are not used in the in-vehicle device 1 are included in the in-vehicle device 1, and the recognition accuracy may be lowered.

  Next, the operation of the voice recognition system will be described. As shown in FIG. 2, the terminal-side control circuit unit 15 includes a terminal-side speech recognition unit 15a, a terminal-side processing control unit 15b, and a dictionary update unit 15c as functional configurations.

  The terminal-side speech recognition unit 15 a performs speech recognition on the speech signal of the uttered speech input from the input device 11 using the terminal-side recognition dictionary in the terminal-side recognition dictionary storage unit 14. If the speech recognition is successful, the terminal-side processing control unit 15b performs processing such as outputting the recognition result word to the output device 12 according to the speech recognition result of the terminal-side speech recognition unit 15a. If it fails, the terminal side communication unit 13 is used to communicate with the center 2 to perform processing such as receiving the speech recognition result of the uttered speech from the center 2. The dictionary updating unit 15c additionally records the vocabulary data received from the center 2 in the terminal-side recognition dictionary, as will be described later, according to the processing result of the terminal-side processing control unit 15b.

  In addition to the functions as described above, the terminal-side control circuit unit 15 accesses the Web server using the terminal-side communication unit 13, receives Web page data, and outputs the output device according to the received Web page data. Processing such as displaying the Web page on 12 displays is also performed.

  Moreover, the center side control circuit part 23 has the center side process control part 23a and the center side speech recognition part 23b as the function structure.

  The center-side processing control unit 23a uses the center-side communication unit 21 to receive data from the vehicle-mounted device 1 and transmit data to the vehicle-mounted device 1 according to the voice recognition result of the center-side voice recognition unit 23b. Do. The center side voice recognition unit 23 b performs voice recognition on the voice feature data of the uttered voice received from the in-vehicle device 1 using the center side recognition dictionary in the center side recognition dictionary storage unit 22.

  FIG. 3 shows a flowchart of processing executed by the terminal-side control circuit unit 15. The specific operation of the speech recognition system will be described below according to this flowchart.

  First, the terminal-side control circuit unit 15 accesses the Web server of a search site (for example, Google (registered trademark)) using the terminal-side communication unit 13, receives data of the Web page of the search site, and receives the received Web It is assumed that the Web page is displayed on the display of the output device 12 according to the page data. At this time, it is assumed that the Web page includes a character input field (input form) for inputting a search word.

  At this time, the user performs an operation for starting voice recognition (for example, pressing a voice recognition start button) to an operation unit (not shown) of the in-vehicle device 1 in order to make the character input in the character input field be voice-recognized. And speak the voice you want to recognize. Hereinafter, the uttered voice is referred to as uttered voice.

  Then, the terminal-side control circuit unit 15 starts the process of FIG. 3 based on the operation for starting the voice recognition, and first executes the speech voice input process in step 110. Specifically, the voice signal of the uttered voice output from the input device 11 is received, and voice feature data based on the received voice signal is acquired. Here, the voice feature data is data representing the feature of the voice signal of the speech voice, for example. For example, it may be feature value data of an audio signal, frequency spectrum data of an audio signal, or intensity data sampled in time series. It is desirable that the data be in the same format as the comparison voice feature data.

  The period for receiving the audio signal may be a predetermined period, a period until the user performs an operation to end the voice recognition, or a period in which the level of the acquired audio signal is lower than the threshold value. It may be a period until it continues for a predetermined period or longer.

  Subsequently, in step 115, the acquired speech feature data and the speech feature data for comparison of each word in the terminal side recognition dictionary recorded in the terminal side recognition dictionary storage unit 14 are converted into a known method (for example, hidden Markov By comparison with a model method), and specifies one piece of comparative voice feature data having the highest likelihood (similarity) with the voice feature data and higher than a predetermined reference value.

  However, since there may be no comparison voice feature data having a likelihood higher than the predetermined reference value, the voice recognition has failed in that case. Conversely, if there is at least one comparison voice feature data whose likelihood with the voice feature data is higher than a predetermined reference value, the voice recognition is successful.

  When the speech recognition is successful, a word corresponding to the extracted comparison speech feature data is extracted as a word corresponding to the uttered speech. Specifically, character string data included in the same vocabulary data as the extracted comparison voice feature data is extracted. The function realized by the processing of these steps 110 and 115 corresponds to the function of the terminal side voice recognition unit 15a.

  Subsequently, in step 120, it is determined whether the speech recognition process in step 115 has succeeded or failed. If it is determined that the speech recognition process has succeeded, the process proceeds to step 125, and the character string of the extracted word (for example, “convenience store”). ) Is output to the output device 12. Thereby, the output device 12 displays the character string of the word in the character input field of the Web page. Thereafter, when the user performs a transmission operation on the operation unit of the in-vehicle device 1 (not shown), the terminal-side control circuit unit 15 transmits the character string as a search word to the Web server of the search site. To do. Then, the Web server transmits data of items that hit the search word (for example, other Web sites) to the in-vehicle device 1, and the terminal-side control circuit unit 15 receives the data and causes the output device 12 to display the data. . The function realized by the processing of steps 120 and 125 corresponds to part of the function of the terminal-side process control unit 15b.

  On the other hand, if it is determined in step 120 that the voice recognition has failed, the process proceeds to step 130 where the voice data based on the uttered voice is used as inquiry voice data and transmitted to the center 2 using the terminal side communication unit 13. Then, in step 135, the process waits until a response to the inquiry voice data is received from the center 2.

  The inquiry voice data may be data representing the characteristics of the speech voice. For example, the inquiry voice data may be data of the feature amount of the voice signal of the speech voice, or may be data of the frequency spectrum of the voice signal. It may be intensity data sampled in time series.

  The center side control circuit unit 23 of the center 2 receives the inquiry voice data transmitted as described above from the vehicle-mounted device 1 through the center side communication unit 21 by the function of the center side processing control unit 23a.

  Moreover, the center side control circuit part 23 performs the following processes by the function of the center side voice recognition part 23b. First, voice feature data is created based on the received inquiry voice data. The voice feature data to be created may be data of the feature amount of the voice signal of the utterance voice, data of the frequency spectrum of the voice signal of the utterance voice, or intensity obtained by sampling the utterance voice in time series However, it is desirable that the data be in the same format as the comparison voice feature data in the center side recognition dictionary. If the inquiry voice data has the same format as the comparison voice feature data in the center recognition dictionary, it may be used as the voice feature data as it is.

  Then, the created voice feature data and the voice feature data for comparison of each word in the center side recognition dictionary recorded in the center side recognition dictionary storage unit 22 are obtained by a known method (for example, a method using a hidden Markov model). ) To identify one piece of comparative voice feature data having the highest likelihood (similarity) with the voice feature data and higher than a predetermined reference value.

  However, as in the process of step 115 in FIG. 3, there may be no comparison voice feature data whose likelihood with the voice feature data is higher than a predetermined reference value. Recognition has failed. Conversely, if there is at least one comparison voice feature data whose likelihood with the voice feature data is higher than a predetermined reference value, the voice recognition is successful.

  When the speech recognition is successful, a word corresponding to the extracted comparison speech feature data is extracted as a word corresponding to the uttered speech. Specifically, vocabulary data including the extracted comparison voice feature data is extracted. The center-side control circuit unit 23 functions as an example of a center-side speech recognition unit by executing such processing that realizes the function of the center-side speech recognition unit 23b.

  Further, the center-side control circuit unit 23 determines whether the word extraction has succeeded or failed by the function of the voice recognition unit 23b by the function of the center-side processing control unit 23a.

  And when it determines with having succeeded, the recognition result containing the vocabulary data extracted by the function of the center side speech recognition part 23b is transmitted to the vehicle equipment 1 using the center side communication part 21. FIG. The vocabulary data includes a character string of a word corresponding to the uttered voice and comparison voice data of the word.

  On the other hand, when it is determined that the word extraction has failed, the recognition result including failure data indicating the failure is transmitted to the in-vehicle device 1 using the center side communication unit 21. The center-side control circuit unit 23 functions as an example of a response unit by executing such a process that realizes the function of the center-side process control unit 23a.

  In step 135, the terminal-side control circuit unit 15 receives the recognition result as a response to the inquiry voice data from the center 2 using the terminal-side communication unit 13b, and then proceeds to step 140.

  In step 140, it is determined whether the voice recognition has succeeded or failed in the center 2 based on the content of the received recognition result. Specifically, if the recognition result includes vocabulary data, it is determined that the speech recognition has succeeded, and if the recognition result includes failure data, it is determined that the speech recognition has failed. When it determines with having failed, it progresses to step 145, the information to the effect that speech recognition has failed is output to the output device 12, and the process of FIG. 3 is complete | finished. Part of the functions of the terminal-side process control unit 15b is realized by the processes of steps 130, 135, and 145.

  If it is determined that the process has succeeded, the process proceeds to step 150, and the vocabulary data in the received recognition result is additionally registered in the terminal-side recognition dictionary of the terminal-side recognition dictionary storage unit 14. That is, the comparison voice feature data included in the recognition result received from the center 2 is additionally registered in the terminal side recognition dictionary as the comparison voice feature data of the word included in the recognition result. The additional registered vocabulary data is vocabulary data that cannot be recognized on the in-vehicle device 1 side, and in most cases, is vocabulary data of words that are not included in the terminal-side recognition dictionary. By the processing in step 150, the function of the dictionary update unit 15c is realized.

  In step 160, the output device 12 is caused to output information based on the character string of the word in the vocabulary data included in the received recognition result. Specifically, as in step 125, the character string of the word (for example, “convenience store”) is output to the output device 12. Thereby, the output device 12 displays the character string of the word in the character input field of the Web page. Thereafter, when the user performs a transmission operation on the operation unit of the in-vehicle device 1 (not shown), the terminal-side control circuit unit 15 transmits the character string as a search word to the Web server of the search site. To do. Then, the Web server transmits data of items that hit the search word (for example, other Web sites) to the in-vehicle device 1, and the terminal-side control circuit unit 15 receives the data and causes the output device 12 to display the data. . The function realized by the process of step 160 corresponds to a part of the function of the terminal-side process control unit 15b.

  As described above, the in-vehicle device 1 selects the case where the user's speech voice has failed to be recognized by the user's own device, along with the word of the speech recognition result of the speech sound from the center 2, in the center side recognition dictionary. The comparison voice feature data for the word is received, and the received comparison voice feature data for the word is added to the terminal side recognition dictionary.

  By doing in this way, the comparison voice feature data of the words that need to be recognized by the speech recognition terminal at least once is selectively registered in the speech recognition terminal. The possibility that such a word needs to be recognized again is higher than a word that does not need to be recognized once, so the comparative voice feature data of the word is relatively likely to be wasted. Low. Therefore, it is possible to reduce the increase in the size of the recognition dictionary in the speech recognition terminal while expanding the recognition dictionary of the speech recognition terminal using the center recognition dictionary.

In the above embodiment, the terminal-side control circuit unit 15 functions as an example of the terminal-side voice recognition unit by executing steps 110 and 115 of FIG. By executing step 150, it functions as an example of a dictionary updating unit.
(Other embodiments)
As mentioned above, although embodiment of this invention was described, the scope of the present invention is not limited only to the said embodiment, The various form which can implement | achieve the function of each invention specific matter of this invention is included. It is.

  For example, in the above embodiment, the word extracted as the word corresponding to the uttered voice is the search word input to the search site, but the word extracted as the word corresponding to the uttered voice is not necessarily such. Not limited to. For example, when the in-vehicle device 1 is a navigation device that has map data including correspondences between names and locations of a plurality of facilities and calculates and guides a guidance route from the current location to the destination, The name of the facility is uttered as uttered speech, and in steps 125 and 160, the destination is searched from the map data using the word character string extracted as the word corresponding to the uttered speech, and the search result is output to the output device 12. You may be made to output to. That is, in steps 125 and 160, the processing result using the search result word may be output to the output device 12. That is, the output device 12 may perform any display as long as the display is based on the search result word. If the utterance voice of the destination name fails in the in-vehicle device 1 and the speech recognition is successful in the center 2, the in-vehicle device 1 receives the vocabulary data from the center 2 and additionally registers it in the terminal side recognition dictionary. Will do. As described above, if the vocabulary data is newly added to the vehicle-mounted device 1, a useful destination name is newly included after the vehicle-mounted device 1 is manufactured, and thus is not included in the above-described correspondence data. The name of the facility can be considered. In such a case, the in-vehicle device 1 may further acquire the location data of the facility having the name from the center 2 and additionally register the location data of the above-described correspondence relationship.

  In the above embodiment, each function realized by the control circuit 17 executing the program is realized by using hardware having those functions (for example, an FPGA capable of programming the circuit configuration). You may come to do.

  In the above embodiment, the terminal-side control circuit unit 15 has realized the functions of the terminal-side voice recognition unit 15a, the terminal-side processing control unit 15b, and the dictionary update unit 15c. 15a, the terminal-side process control unit 15b, and the dictionary update unit 15c may be realized as separate ICs.

  Similarly, in the above embodiment, the center-side control circuit unit 23 is configured to realize the functions of the center-side processing control unit 23a and the center-side voice recognition unit 23b. The voice recognition unit 23b may be realized as a separate IC.

  Moreover, in the said embodiment, although the vehicle equipment 1 is used as an example of a voice recognition terminal, a voice recognition terminal does not necessarily need to be a vehicle-mounted apparatus. For example, a terminal carried by the user may be used.

DESCRIPTION OF SYMBOLS 1 Vehicle equipment 2 Center 11 Input device 12 Output device 13 Terminal side communication part 14 Terminal side recognition dictionary memory | storage part 15 Terminal side control circuit part 15a Terminal side speech recognition part 15b Terminal side processing control part 15c Dictionary update part 21 Center side communication part 22 Center side recognition dictionary storage unit 23 Center side control circuit unit 23a Center side processing control unit 23b Center side speech recognition unit

Claims (3)

A speech recognition system comprising a speech recognition terminal (1) and a center (2),
The voice recognition terminal (1) includes a vehicle-side recognition dictionary storage unit (14) in which a vehicle-side recognition dictionary is recorded, and a terminal-side control circuit unit (15).
The center (2) includes a center side recognition dictionary storage unit (22) in which a center side recognition dictionary is recorded, and a center side control circuit unit (23).
The center side recognition dictionary has voice feature data for comparison of words that the terminal side recognition dictionary does not have,
The terminal side control circuit unit (15)
Voice feature data based on the user's spoken voice is acquired (110), and based on a comparison between the acquired voice feature data and the voice feature data for comparison of each word in the vehicle-side recognition dictionary, it corresponds to the spoken voice Terminal-side speech recognition means (110, 115) for extracting words to be
Inquiry transmission means for transmitting voice data based on the uttered voice to the center (2) as inquiry voice data based on the fact that the terminal side voice recognition means (15a, 110, 115) has failed to extract a word. 130), and
The center side control circuit section (23)
Corresponds to the uttered speech based on a comparison between speech feature data based on the inquiry speech data transmitted from the speech recognition terminal (1) and comparison speech feature data for each word in the center-side recognition dictionary. Center side speech recognition means (23b) for extracting words;
Response means for transmitting a recognition result including the word extracted by the center side speech recognition means (23b) and comparison voice feature data of the word in the center side recognition dictionary to the voice recognition terminal (1). (23a)
Further, the terminal-side control circuit unit (15) uses the comparison voice feature data included in the recognition result received from the center (2) as comparison voice feature data of words included in the received recognition result. A speech recognition system comprising dictionary updating means (150) for additionally registering in the terminal side recognition dictionary. A speech recognition terminal that communicates with the center (2),
An in-vehicle side recognition dictionary storage unit (14) in which an in-vehicle side recognition dictionary is recorded, and a terminal side control circuit unit (15),
The terminal side control circuit unit (15)
Voice feature data based on the user's spoken voice is acquired (110), and based on a comparison between the acquired voice feature data and the voice feature data for comparison of each word in the vehicle-side recognition dictionary, it corresponds to the spoken voice Terminal-side speech recognition means (110, 115) for extracting words to be
Inquiry transmission means for transmitting voice data based on the uttered voice to the center (2) as inquiry voice data based on the fact that the terminal side voice recognition means (15a, 110, 115) has failed to extract a word. 130),
Based on the comparison between the voice feature data based on the inquiry voice data transmitted from the voice recognition terminal (1) and the voice feature data for comparison of each word in the center side recognition dictionary, the center (2) When a word corresponding to the uttered voice is extracted, and a recognition result including the extracted word and voice feature data for comparison of the word in the center-side recognition dictionary is transmitted to the voice recognition terminal, the center ( Dictionary update means (150) for additionally registering the comparison speech feature data included in the recognition result received from 2) as the comparison speech feature data of the word included in the received recognition result in the terminal side recognition dictionary And a voice recognition terminal. A center that communicates with the voice recognition terminal (1),
A center side recognition dictionary storage unit (22) in which a center side recognition dictionary is recorded, and a center side control circuit unit (23),
The center side control circuit section (23)
When the voice recognition terminal (1) transmits voice data based on the user's uttered voice to the center (2) as inquiry voice data, the voice feature data based on the inquiry voice data and the center side recognition dictionary Center-side speech recognition means (23b) for extracting a word corresponding to the uttered speech based on the comparison of each word with comparison speech feature data;
A recognition result including the word extracted by the center side speech recognition means (23b) and comparison voice feature data of the word in the center side recognition dictionary is transmitted to the speech recognition terminal (1), The response means for causing the voice recognition terminal (1) to additionally register the comparison voice feature data included in the recognition result as the comparison voice feature data of the word included in the recognition result in the terminal side recognition dictionary. (23a).

Images