JP4941494B2 - Speech recognition system - Google Patents

Description

  The present invention relates to a speech recognition system applied to, for example, a vehicle navigation apparatus.

  In recent years, a system for recognizing a voice spoken by a user and operating an operation target device or providing a service to the user based on the recognition result has been developed and put into practical use.

  For example, Patent Document 1 describes a voice response device that recognizes a user's voice and provides services such as order reception and database search based on the recognition result. In this voice response device, the voice recognition unit recognizes which words and phrases previously registered in the voice recognition dictionary unit are uttered in what order. Then, the unnecessary word detection unit checks whether or not the recognition result of the voice recognition unit includes an unnecessary word that is a phrase that is not necessary for the operation of the voice response device. When an unnecessary word is included, the unnecessary word detection unit further checks the positional relationship between the unnecessary word and the target word in the recognition result.

  Then, the proficiency level estimation unit estimates the proficiency level of the user's operation of the voice response device based on the result of the examination by the unnecessary word detection unit. The conversation flow control unit extracts the guidance included in the conversation flow corresponding to the estimated proficiency level from the conversation flow stored in advance, and outputs the guidance to the user.

JP 2001-331196 A

  In the voice response device described in Patent Document 1, as described above, guidance according to the proficiency level of the user's operation is performed.

  However, as described in Patent Document 1, if an attempt is made to estimate the proficiency level of the user's operation from whether or not unnecessary words are included and the positional relationship between the unnecessary words and the target word, a huge recognition vocabulary is required. It is necessary to use a speech recognition dictionary consisting of In other words, in addition to the recognition vocabulary for recognizing an enormous number of target words, recognition vocabularies for recognizing a large number of unnecessary words need to be collated with the user's speech. When collation with such a large number of recognition vocabularies is performed, the probability of misrecognition increases on the contrary, and voice recognition performance may be degraded.

  The present invention has been made in view of these points, and suppresses a decrease in speech recognition performance by performing speech recognition using a dictionary with a small number of unnecessary words included in the recognition vocabulary as much as possible. An object of the present invention is to provide a voice recognition system capable of

In order to achieve the above object, a speech recognition system according to claim 1 comprises:
A voice input unit for inputting voice;
Speaker identification means for identifying a speaker based on the voice input to the voice input unit;
A speech recognition unit for recognizing speech input to the speech input unit using a dictionary having a large number of recognition vocabularies;
For each speaker identified by the speaker identification means, unnecessary words that are originally unnecessary as input speech in the speech recognized by the speech recognition unit are counted, and the unnecessary word usage frequency calculated based on the count result is calculated. An unnecessary word use frequency storage unit for storing;
Multiple dictionaries with different numbers of unnecessary words are prepared as dictionaries to be used in the speech recognition unit, and unnecessary words are stored in the unnecessary word usage frequency storage unit for speakers identified by the speaker identification means. And a dictionary switching unit that switches the use dictionary in the speech recognition unit to a dictionary having the number of unnecessary words corresponding to the stored unnecessary word use frequency when the use frequency is stored.

  As described above, according to the first aspect of the present invention, the speaker is identified from the voice input to the voice input unit, and the use frequency of the unnecessary word in the voice of the speaker is calculated. It is stored as an unnecessary word usage frequency. This unnecessary word usage frequency represents a tendency of how often unnecessary words are used for each speaker.

  Therefore, when a speaker is identified on the basis of the input speech, and the unnecessary word usage frequency is stored for the speaker, the dictionary used in the speech recognition unit is stored as a stored unnecessary word. Switch to a dictionary with as many unnecessary words as you want. As a result, a dictionary with a relatively large number of unnecessary words is used for voice recognition of users who frequently use unnecessary words, but it is not necessary for voice recognition of users who use less unnecessary words. A dictionary with a relatively small number of words is used. Thus, according to the first aspect of the present invention, as the dictionary for speech recognition, a dictionary having the number of unnecessary words corresponding to the frequency of use of unnecessary words by the user is used. A dictionary with few can be used.

  As described in claim 2, the dictionary includes an object dictionary that collects objects that are vocabulary required as input speech, and an unnecessary word dictionary that collects unnecessary words that are originally unnecessary as input speech. A plurality of dictionaries with different numbers of unnecessary words are prepared as the unnecessary word dictionaries, and the dictionary switching unit switches the unnecessary word dictionaries according to the stored unnecessary word use frequency, and is used for speech recognition. The unit preferably performs speech recognition using the target word dictionary and the unnecessary word dictionary switched by the dictionary switching unit. In this way, by separating the target word dictionary and the unnecessary word dictionary and making only the unnecessary word dictionary a switching target, it is possible to prevent the dictionary capacity from becoming excessive.

  As described in claim 3, it is preferable that the dictionary switching unit can switch the unnecessary word dictionary so that the unnecessary word dictionary is not used. For example, if the user is familiar with the operation of the voice recognition system, an unnecessary word dictionary may not be required for the recognition of the input voice.

  According to a fourth aspect of the present invention, there is provided a determining unit that determines a type of information input by voice based on a user operation, and the dictionary switching unit is configured such that the type of the voice input information is determined by the determining unit In consideration of the determined type of voice input information, it is preferable to switch the use dictionary in the voice recognition unit.

  For example, when the vehicle navigation device is an operation target device and the information for setting the destination is a combination of geographical information and a genre, and a case where a phone number is input, the geographical information is The unnecessary words tend to increase when the genre is input by voice. As described above, the type of information input by speech and the number of unnecessary words included in the input speech have a certain degree of correlation. Therefore, as described in claim 4, it is preferable to switch the use dictionary in consideration of the type of voice input information.

  Specific methods for switching the use dictionary in consideration of the type of voice input information are described in claims 5 and 6.

  That is, as described in claim 5, when the type of the voice input information is determined by the determining unit, the dictionary switching unit determines the use regardless of the use frequency of unnecessary words of the speaker identified by the speaker identifying unit. The dictionary may be switched to the number of unnecessary words corresponding to the type of the voice input information. This is because the type of voice input information is considered to have a greater influence on the number of unnecessary words included in the user's uttered voice than the individual difference for each user.

  Further, as described in claim 6, the dictionary switching unit responds to the dictionary of the number of unnecessary words according to the frequency of unnecessary word usage of the speaker identified by the speaker identifying means and the type of the voice input information. A dictionary having a larger number of unnecessary words may be switched to a dictionary having the number of unnecessary words. Thereby, the dictionary used in the speech recognition unit can be switched to a dictionary appropriately corresponding to the number of unnecessary words used due to individual differences for each user and the number of unnecessary words used due to the type of voice input information. It becomes possible.

It is a block diagram showing the structure of the navigation apparatus for vehicles provided with the speech recognition system by embodiment of this invention. It is a control block diagram which shows the detailed structure of a speech recognition part and a dialogue control part in a speech recognition system. (a) is explanatory drawing showing a mode that the frequency of the unnecessary word was memorize | stored for every user, (b) is explanatory drawing which shows an example of the reference | standard at the time of selecting an unnecessary word dictionary according to an unnecessary word frequency. . It is a flowchart which shows the main control processing in a speech recognition system.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the embodiment described below, the voice recognition system of the present invention is applied to a vehicle navigation apparatus, but the application target of the voice recognition system of the present invention is not limited to the vehicle navigation apparatus.

  As shown in FIG. 1, the vehicle navigation device 2 includes a position detector 4, a data input device 6, an operation switch group 8, a control circuit 10 connected thereto, a communication device 12 connected to the control circuit 10, and an external device. A memory 14, a display device 16, a remote control sensor 18, and a voice recognition system 30 are provided. The control circuit 10 is configured as a normal computer, and includes a well-known CPU, ROM, RAM, I / O, and a bus line for connecting these configurations.

  The position detector 4 has a known gyroscope 20, a distance sensor 22, and a GPS receiver 24 for detecting the position of the vehicle based on radio waves from a satellite. Since these sensors 20, 22, and 24 have errors of different properties, they are configured to be used while being interpolated by a plurality of sensors. Depending on the accuracy, a part of the above may be used, and further, a steering rotation sensor, a wheel sensor of each rolling wheel, or the like may be used.

  In addition to so-called map matching data for improving the accuracy of position detection, various data for navigation including map data and landmark data, the data input device 6 uses dictionary data used when the speech recognition system 30 performs recognition processing. Is a device for inputting. As a storage medium, a hard disk or a DVD is generally used because of its data amount, but another medium such as a CD-ROM may be used.

  For example, a touch switch or a mechanical switch integrated with the display device 16 is used as the operation switch group 8, and various operation instructions are output to the control circuit 10 by the switch operation. For example, operation instructions such as map scale change, menu display selection, destination setting, route search, route guidance start, display screen change, voice guidance setting, and volume adjustment are performed. Further, the operation switch group 8 includes, for example, switches for selecting the type of information for setting the departure place and the destination. By operating the selection switch, the user (vehicle occupant) sets the starting point and the destination using desired information such as a registered point, facility name, telephone number, and address. Can do.

  The communication device 12 is for communicating with a contact specified by the set contact communication information, and is constituted by a mobile communication device such as a mobile phone. The external memory 14 is a writable mass storage device. The external memory 14 stores a large amount of data and data that cannot be deleted even when the power is turned off, and frequently used data is copied from the map data input device 6 and used. . The external memory 14 may be a removable memory having a relatively small storage capacity.

  The display device 16 is composed of, for example, a liquid crystal display device. On the screen of the display device 16, a vehicle current position mark indicating the current position of the vehicle detected by the position detector 4 and the vehicle input from the map data input device 6 are displayed. Map data around the current position and additional data such as a guide route and a set point mark displayed on the map can be displayed in an overlapping manner. In addition, when a menu screen that displays a plurality of options, or when an option is selected, a command input screen that displays a plurality of options can be displayed.

  The remote control sensor 18 receives an operation signal from a remote controller (not shown) and outputs it to the control circuit 10. The remote controller is provided with a large number of switches, and by operating the switches, it is possible to instruct the control circuit 10 to execute a function substantially equivalent to the operation switch group 8.

  Next, the configuration of the speech recognition system 30 will be described with reference to FIGS. 1 and 2. FIG. 2 is a block diagram showing a detailed configuration of the voice recognition unit 31 and the dialogue control unit 32 in the voice recognition system 30.

  The voice recognition system 30 is configured to allow various commands to be input to the control circuit 10 even by the user's uttered voice, while the operation switch group 8 or the remote controller is manually operated for inputting various commands. It is.

  The voice recognition system 30 includes a voice recognition unit 31, a dialogue control unit 32, a voice synthesis unit 33, a voice extraction unit 34, a microphone 35, a talk switch 36, a speaker 37, and a control unit 38.

  The talk switch 36 is used to instruct the user (driver) to start voice input, and is provided at a position where the user can easily operate, for example, near the side surface of the steering column cover or the shift lever. Note that the talk switch 36 is a so-called click-type switch, and after the user turns on the talk switch 36, voice is input (speaking). Based on the input of the ON signal from the talk switch 36, the control unit 38 instructs the audio extraction unit 34 to execute an audio signal extraction process. In addition, the control unit 38 notifies the voice recognition unit 31 and the dialogue control unit 32 that the voice extraction processing in the voice extraction unit 34 has started. Then, the dialogue control unit 32 outputs a guidance voice such as “Please input voice” from the speaker 37 via the voice synthesis unit 33. The voice synthesizer 33 synthesizes a voice based on a response voice output instruction from the dialogue control unit 32 using a voice waveform stored in the waveform database. This synthesized voice is output from the speaker 37.

  The microphone 35 to which the user's speech is input is provided at a position where the user's voice can be easily picked up, such as an upper surface portion of the steering column cover or a sun visor on the driver's seat side. The voice extraction unit 34 takes a voice signal from the microphone 35 in accordance with an instruction from the control unit 38, and removes a noise component from the voice signal to extract voice data. The extracted voice data is output to the voice recognition unit 31.

  The process in the voice extraction unit 34 will be described in a little more detail. The voice extraction unit 34 converts surrounding voice signals captured by the microphone 35 into digital voice data. Then, for example, a frame signal in a section of about several tens of ms is cut out at regular intervals, and it is determined whether the input signal is a voice section in which voice is included or a noise section in which voice is not included. Since the signal input from the microphone 35 includes not only the speech to be recognized but also noise, the speech section and the noise section are determined. Many methods have been proposed as this determination method. For example, the short-time power of the input signal is extracted at regular intervals, and depending on whether or not the short-time power equal to or greater than a predetermined threshold continues for a certain period. A method of determining whether a voice section or a noise section is often used. And when it determines with it being an audio | voice area, it outputs it to the audio | voice recognition part 31 as audio | speech data.

  The voice recognition unit 31 includes an extraction result storage unit 311, a collation unit 312, and a dictionary unit 313 as shown in FIG. 2.

  The extraction result storage unit 311 stores the audio data from which the noise component input from the audio extraction unit 34 has been removed. The collation unit 312 collates the speech data stored in the extraction result storage unit 311 using the object word dictionary 313a and the unnecessary word dictionary 313b in the dictionary unit 313 (recognition process), and compares a plurality of comparison targets. The higher comparison target pattern having a higher degree of matching than the pattern candidate is output to the dialogue control unit 32 as a voice recognition result.

  In this embodiment, as shown in FIG. 2, the dictionary unit 313 includes an unnecessary word dictionary 313b in addition to the target word dictionary 313a. The target word dictionary 313a is a dictionary that collects target words that are words necessary for voice operations such as commands and destinations, and the unnecessary word dictionary 313b is a dictionary that collects unnecessary words that are words that are not necessary for voice operations. It is. In the present embodiment, the unnecessary word dictionary 313b includes an unnecessary word (large) dictionary 313ba having a large number of unnecessary words and an unnecessary word (small) dictionary 313bb having a small number of unnecessary words. The dictionary unit 313 can switch the unnecessary word dictionary 313b used by the matching unit 312 in accordance with an instruction from the dictionary switching unit 326 of the dialogue control unit 32 described later. Furthermore, when the non-use of the unnecessary word dictionary 313b is instructed from the dictionary switching unit 326, it is possible to provide only the target word dictionary 313a without providing the unnecessary word dictionary 313b to the matching unit 312. That is, in this case, the collation unit 312 performs the above-described recognition process using only the target word dictionary 313a in a state where there is no unnecessary word dictionary.

  As shown in FIG. 2, the dialogue control unit 32 includes a processing unit 321, a speaker specifying unit 322, an unnecessary word usage frequency counting unit 323, an unnecessary word usage frequency storage unit 324, an unnecessary word usage frequency determining unit 325, and a dictionary switching. Part 326.

  Based on the recognition result in the voice recognition unit 31 and the instruction from the control unit 38, the processing unit 321 performs an instruction to output a response voice to the voice synthesis unit 33 or the control circuit 10 that executes the processing of the navigation device 2 itself. Then, a recognition result in the voice recognition unit 31, for example, a destination or a command is notified, and processing for instructing to set the destination or execute the command is performed. As a result of such processing, if the voice recognition system 30 is used, a destination can be instructed to the navigation device 2 by voice input without manually operating the operation switch group 8 or the remote controller.

  In addition, when the control circuit 10 determines the type of information for setting the destination, for example, by the manual operation of the operation switch group 8 or the remote controller or the voice operation, the processing unit 321 is determined. The type of information is notified from the control circuit 10. Then, the processing unit 321 instructs the dictionary switching unit 326 to select the unnecessary word dictionary 313b having the number of unnecessary words corresponding to the determined type of information. Thus, when the determined type of information is input by voice, the input voice can be recognized using a dictionary of unnecessary words suitable for the type of information.

  For example, as the information for setting the destination, the geographical information and the genre are used for voice input (for example, “Ramen shop near Nagoya Station”) and the telephone number is used for voice input. The number of unnecessary words tends to increase when information and genre are input by voice. As described above, the type of information input by speech and the number of unnecessary words included in the input speech have a certain degree of correlation. Therefore, by performing switching of unnecessary word dictionaries based on the type of voice input information, it is possible to perform voice recognition processing using an unnecessary word dictionary including an appropriate number of unnecessary words.

  In order to instruct the control circuit 10 the type of information for setting the destination by voice operation, the type of information (“address”, “ Just say “facility name”, “genre” (including geographical information), “phone number”, etc.

  The speaker specifying unit 322 inputs the voice data stored in the extraction result storage unit 311 of the voice extracting unit 34, and specifies a user as a speaker based on the voice data. That is, the speaker specifying unit 322 generates and stores a so-called speaker model indicating the acoustic characteristics of the speech uttered by each user for each user, and uses the speaker model to make the utterance from the actual utterance. Identify the Lord. At this time, the comparison process between the speaker model and the acoustic feature of the utterance, for example, calculates the similarity between the speaker model and the acoustic feature of the utterance, and uses the calculated similarity as the speaker model. Compare with the corresponding similarity threshold, and when the difference between the similarity and the threshold is within a predetermined range, specify that the utterance is an utterance by the speaker corresponding to the speaker model be able to.

  The unnecessary word usage frequency counting unit 323 counts the usage frequency of unnecessary words in the entire recognized word / phrase when a voice actually spoken by a speaker is recognized by the matching unit 312. Then, as shown in FIG. 3A, the count result is stored as it is in the unnecessary word use frequency storage unit 324 for each speaker specified by the speaker specifying unit 322, or the count result The frequency of unnecessary word usage updated using is stored. That is, when the unnecessary word usage frequency is not stored in the unnecessary word usage frequency storage unit 324 corresponding to the speaker specified by the speaker specifying unit 322, the count result by the unnecessary word usage frequency counting unit 323 is displayed. The unnecessary word use frequency storage unit 324 is stored as it is. On the other hand, if the unnecessary word usage frequency corresponding to the speaker specified by the speaker specifying unit 322 is already stored in the unnecessary word usage frequency storage unit 324, the unnecessary word usage frequency counting unit counts this time. The unnecessary word usage frequency updated to reflect the result obtained is stored.

  If the speaker specification unit 322 identifies the speaker, the unnecessary word usage frequency determination unit 325 may store the unnecessary word usage frequency corresponding to the speaker if the unnecessary word usage frequency storage unit 324 stores the unnecessary word usage frequency. In order to input the stored unnecessary word usage frequency and recognize the utterance voice of the speaker, it is determined how many unnecessary word dictionaries including unnecessary words are necessary.

  For example, as shown in FIG. 3B, if the frequency of use of unnecessary words is 50% or more, the speaker thinks that the use frequency of unnecessary words is high and there are many types of unnecessary words to be used. Therefore, it is determined that the unnecessary word (large) dictionary 313ba having the largest number of unnecessary words is necessary. Also, if the frequency of use of unnecessary words is greater than 0% and less than 50%, the use frequency of unnecessary words is not so high, and the types of unnecessary words used are considered to be limited. The unnecessary word (small) dictionary 313bb having a relatively small number is determined to be appropriate. If the use frequency of the unnecessary word is 0%, it is proficient in voice operation and it is considered that the speaker speaks only the object without using the unnecessary word, so the unnecessary word dictionary is unnecessary. Is determined.

  The dictionary switching unit 326 is an unnecessary word used when the collation unit 312 performs input speech recognition processing based on an unnecessary word dictionary selection instruction from the processing unit 321 and a determination result of the unnecessary word use frequency determination unit 325. Switch dictionary 313b.

  For example, the dictionary switching unit 326 selects from the processing unit 321 the unnecessary word dictionary 313b having the number of unnecessary words corresponding to the determined type of information based on the determined type of input information. When instructed, the unnecessary word dictionary 313b instructed by the processing unit 321 is switched regardless of the determination result of the unnecessary word use frequency determining unit 325. On the other hand, when there is no instruction regarding the selection of the unnecessary word dictionary 313b from the processing unit 321, the unnecessary word dictionary 313b is switched based on the determination result of the unnecessary word usage frequency determination unit 325. The reason for this is that the type of voice input information is considered to have a larger influence on the number of unnecessary words included in the user's uttered voice than the individual difference for each user.

  Next, main control processing including switching processing of the unnecessary word dictionary 313b in the voice recognition system 30 described above will be described based on the flowchart of FIG.

  First, in step S110, it is determined whether or not the talk switch 36 is turned on. At this time, if it is determined that the talk switch 36 is turned on, the process proceeds to step S120. In step S120, voice input processing is performed. That is, the audio extraction unit 34 generates audio data from which noise components have been removed from the audio signal input to the microphone 35.

  In step S130, it is determined whether or not the control circuit 10 notifies that the type of input information has been determined. When the type of input information is determined, the process proceeds to step S140, and the unnecessary word dictionary 313b having the number of unnecessary words corresponding to the determined type of information is selected. Thus, when the determined type of information is input by voice, the number of unnecessary word dictionaries 313b suitable for the type of information is used (including the nonuse of the unnecessary word dictionary 313b). Input speech can be recognized. On the other hand, if it is determined in step S130 that the type of input information has not been determined, the process proceeds to step S150.

  In step S150, the speaker identification unit 322 identifies a speaker based on the extracted voice data. That is, a so-called speaker model indicating the acoustic characteristics of the speech uttered by each user is generated and stored for each user, and the speaker is identified from the actual utterance using the speaker model.

  In step S160, it is determined whether or not an unnecessary word usage frequency corresponding to the specified speaker is stored. If the unnecessary word usage frequency corresponding to the identified speaker is stored, the process proceeds to step S170. In step S170, the number of unnecessary word dictionaries 313b most suitable for recognizing the uttered speech of the identified speaker is determined based on the stored unnecessary word usage frequency (the number of unnecessary word dictionaries 313b). Including non-use). On the other hand, if it is determined in step S160 that the unnecessary word usage frequency is not stored, the process proceeds to step S180. In step S180, the unnecessary word dictionary 313b to be used is determined to be the unnecessary word (large) dictionary 313ba having the largest number of unnecessary words included. This is because the frequency of unnecessary words used by the user who is a speaker is unknown, so that the input speech can be recognized even when unnecessary words are used frequently. .

  In subsequent step S190, recognition processing of the voice input by the user is executed using the unnecessary word dictionary 313b selected in step S140 or determined in step S170 or S180 and the target word dictionary 313a. The recognition result by this recognition processing is output to the control circuit 10 in step S200.

  In step S210, the frequency of use of unnecessary words in the entire recognized word / phrase is counted based on the recognition result. In step S220, the unnecessary word usage frequency count result is stored as it is for each speaker specified in step S150, or the unnecessary word usage frequency updated using the count result is stored. The

  The preferred embodiments of the present invention have been described above, but the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the present invention.

  For example, in the above-described embodiment, when it is notified that the type of information input from the control circuit 10 has been determined, the unnecessary word dictionary 313b is used as the type of input information regardless of the speaker who performs voice input. The number of unnecessary word dictionaries 313b corresponding to the number of unnecessary words is selected.

  However, for example, an unnecessary word dictionary 313b having the number of unnecessary words corresponding to the type of input information and an unnecessary word dictionary 313b having the number of unnecessary words corresponding to the frequency of unnecessary word usage of the specified speaker are respectively obtained. The unnecessary word dictionary 313b may be switched to the unnecessary word dictionary 313b having a larger number of unnecessary words. Accordingly, the unnecessary word dictionary 313b used in the speech recognition unit 31 appropriately corresponds to both the number of unnecessary words used due to individual differences for each user and the number of unnecessary words used due to the type of input information. It becomes possible to switch to the unnecessary word dictionary 313b.

  In the embodiment described above, the unnecessary word dictionary 313b is switched to any one of the three types of the unnecessary word dictionary 313b, the unnecessary word (small) dictionary 313bb, and the unnecessary word (large) dictionary 313ba. The number of unnecessary word dictionaries 313b to be switched may be two or four or more.

30 speech recognition system 31 speech recognition unit 32 dialogue control unit 33 speech synthesis unit 34 speech extraction unit 35 microphone 36 talk switch 37 speaker 38 control unit

Claims (6)

A voice input unit for inputting voice;
Speaker identification means for identifying a speaker based on the voice input to the voice input unit;
A speech recognition unit for recognizing speech input to the speech input unit using a dictionary having a large number of recognition vocabularies;
For each speaker identified by the speaker identifying means, unnecessary words that are unnecessary as input speech in the speech recognized by the speech recognition unit are counted, and the unnecessary word usage frequency calculated based on the count result An unnecessary word usage frequency storage unit for storing
A plurality of dictionaries with different numbers of unnecessary words included in the recognition vocabulary are prepared as the dictionary used in the speech recognition unit, and the use of the unnecessary words for the speakers identified by the speaker identification means A dictionary switching unit that switches the use dictionary in the voice recognition unit to a dictionary of the number of unnecessary words corresponding to the stored unnecessary word use frequency when the unnecessary word use frequency is stored in the frequency storage unit. A speech recognition system characterized by that. The dictionary is composed of an object dictionary that collects objects that are vocabulary required as input speech, and an unnecessary word dictionary that collects unnecessary words that are originally unnecessary as input speech. There are several dictionaries with different numbers.
The dictionary switching unit switches the unnecessary word dictionary according to the stored unnecessary word usage frequency,
The speech recognition system according to claim 1, wherein the speech recognition unit performs speech recognition using the object word dictionary and an unnecessary word dictionary switched by the dictionary switching unit.   The speech recognition system according to claim 2, wherein the dictionary switching unit can switch the unnecessary word dictionary so that the unnecessary word dictionary is not used. A determination unit configured to determine a type of information input by voice based on a user operation;
The dictionary switching unit, when the type of voice input information is determined by the determining unit, switches the dictionary used in the voice recognition unit in consideration of the type of the determined voice input information. The speech recognition system according to any one of claims 1 to 3.   When the type of speech input information is determined by the determining unit, the dictionary switching unit determines the type of the speech input information determined regardless of the frequency of use of unnecessary words of the speaker identified by the speaker identifying unit. 5. The speech recognition system according to claim 4, wherein the dictionary is switched to a dictionary having the number of unnecessary words corresponding to the dictionary.   The dictionary switching unit includes a dictionary of the number of unnecessary words according to the frequency of use of unnecessary words of the speaker identified by the speaker identifying means, and a dictionary of the number of unnecessary words according to the type of the voice input information. 5. The voice recognition system according to claim 4, wherein the dictionary is switched to a dictionary having a larger number of unnecessary words.

Images