JP2021085974A - Speech recognition apparatus, speech recognition processing method and speech recognition processing program - Google Patents

Abstract

To provide a speech recognition apparatus, a speech recognition processing method and a speech recognition processing program that can reduce false recognition.SOLUTION: An on-vehicle apparatus 1 comprises: a microphone 30 which picks up a user's utterance speech; a speech recognition dictionary in which a plurality of words to be selected are registered; a speech recognition processing part 52 which performs feature extraction on the utterance speech picked up by the microphone 30, and selects a word higher in similarity than a predetermined threshold from a plurality of words registered in the speech recognition dictionary; and a recognition result selection part 53 which selects, when there are a plurality of words determined by the speech recognition processing part 52 to be high in similarity, a word having top priority out of them.SELECTED DRAWING: Figure 1

Description

The present invention relates to a voice recognition device, a voice recognition processing method, and a voice recognition processing program that perform voice recognition for a voice uttered by a user.

Conventionally, there has been known an imaging device that sets the priority of functions that are likely to be used according to the state of the camera and updates the command weights of the functions with high priority during voice recognition ( For example, see Patent Document 1.). For example, in the voice recognition of "cheese" included in the function of the "shutter button", it is determined that this indicates the shooting command "cheese" even when it is recognized as "cheese". In this way, the higher the weight of the shooting command, the easier it is to determine that the command is input by the user, even though the similarity with the word obtained by voice recognition is low. ..

Japanese Unexamined Patent Publication No. 2014-122978

By the way, in the voice recognition shown in Patent Document 1 described above, a word having a high priority is easily obtained as a recognition result, but a plurality of similar words that are easily erroneously recognized are included, and one of them is uttered. In such a case, there is a problem that the occurrence of erroneous recognition cannot be suppressed. For example, when the field (product) of voice recognition is expanded to other than the imaging device, any one of 10 words from "1 channel" to "10 channels" included in the channel selection command can be recognized by voice recognition. Shall be selected. When the user actually utters "Icchanneru", if it is judged that the similarity is high in the order of "8 channels", "1 channel", and "10 channels", the recognition result is "8 channels". "Become. Even if the weights of "8 channels", "1 channel", and "10 channels" are increased by the voice recognition method shown in Patent Document 1, such erroneous recognition is unavoidable. If this erroneous recognition is caused by the habit of the user's utterance, the same recognition result will be obtained no matter how many times the utterance is repeated.

The present invention has been created in view of these respects, and an object of the present invention is to provide a voice recognition device, a voice recognition processing method, and a voice recognition processing program capable of reducing erroneous recognition.

In order to solve the above-mentioned problems, the voice recognition device of the present invention includes a sound collecting means for collecting the uttered voice of the user, a voice recognition dictionary in which a plurality of words to be selected are registered, and a sound collecting means. A voice recognition means that extracts features from the spoken voice collected by the voice recognition dictionary and selects a word having a higher degree of similarity than a predetermined threshold value from a plurality of words registered in the voice recognition dictionary, and voice recognition. When there are a plurality of words determined to have high similarity by the means, the recognition result selection means for selecting the word having the highest priority from among them is provided.

Further, the voice recognition processing method of the present invention extracts features of the spoken voice collected by the sound collecting means, and is more similar than a predetermined threshold value from a plurality of words registered in the voice recognition dictionary. When there are a voice recognition step in which a word having a high degree is selected by a voice recognition means and a plurality of words judged to have a high degree of similarity by the voice recognition means, the word having the highest priority is recognized by the recognition result selection means. It has a recognition result selection step to be selected by.

Further, the voice recognition processing program of the present invention extracts features of the spoken voice collected by the sound collecting means by the computer, and sets a predetermined threshold value from a plurality of words registered in the voice recognition dictionary. When there are multiple voice recognition means that select words with a higher degree of similarity and words that are determined to have a higher degree of similarity by the voice recognition means, the recognition result selection that selects the word with the highest priority from among them. Make it function as a means.

When multiple words are obtained as a voice recognition result, an error in voice recognition is made by selecting one of them according to the priority set assuming the order in which the user is likely to select. Recognition can be reduced.

Further, the above-mentioned recognition result selection means is used when there are a plurality of words determined to have high similarity by the voice recognition means and the difference in similarity between the plurality of words is within a predetermined value. It is desirable to select the word with the highest priority. When the difference in similarity between multiple words is small, there is a high possibility that misrecognition will occur. In such a case, by selecting words based on the priority, it is necessary to reduce misrecognition in speech recognition. Can be done.

Further, it is desirable that the above-mentioned priority is set based on the hobbies and tastes of the user who speaks. Alternatively, it is desirable that the above-mentioned priority is set based on the past history of the voice recognition result. This makes it possible to set the priority of words in consideration of the possibility that the user speaks.

Further, it is desirable that the above-mentioned priority is set based on the situation of the user when performing the voice recognition process. As a result, it is possible to set the priority of words in consideration of the possibility that the user speaks according to the operation content of the user during the voice recognition process and the surrounding environment.

Further, it is desirable that the above-mentioned priority is set at the time when the spoken voice to be recognized by the voice is collected by the sound collecting means. By using the latest priority, it is possible to further reduce erroneous recognition in speech recognition.

It is a figure which shows the structure of the vehicle-mounted device of one Embodiment. It is a figure which shows the relationship between the threshold value of a voice recognition process, and a score. It is a flow chart which shows the operation procedure which performs the voice recognition processing in the state which the menu screen corresponding to various commands to be voice recognition is displayed in the in-vehicle device. It is a figure which shows the relationship between the channel designation command (word) at the time of designating a desired television broadcast (channel) using a TV tuner processing unit, and the priority of each channel. It is a figure which shows an example of the search result screen which extracted the nearby convenience store by using the navigation processing part. It is a figure which shows the relationship between the designated command (word) at the time of designating the convenience store to go to in the search result screen shown in FIG. 5, and the priority of each word.

Hereinafter, an in-vehicle device according to an embodiment to which the voice recognition device of the present invention is applied will be described with reference to the drawings.

FIG. 1 is a diagram showing a configuration of an in-vehicle device according to an embodiment. As shown in FIG. 1, the in-vehicle device 1 includes a navigation processing unit 10, a TV tuner processing unit 14, a radio tuner processing unit 16, an AV processing unit 18, an operation unit 20, an input control unit 22, a display processing unit 24, and a display device. It includes 26, a microphone 30, an analog-to-digital converter (A / D) 32, a digital-to-analog converter (D / A) 40, a speaker 42, a control unit 50, and a hard disk device (HDD) 70.

The navigation processing unit 10 uses the map data stored in the hard disk device 70 to perform a navigation operation for guiding the traveling of the vehicle on which the in-vehicle device 1 is mounted. Used together with the GPS (Global Positioning System) device 12 that detects the position of the own vehicle, the navigation operation that guides the vehicle's travel includes map display, route search / guidance, and operation to search and display surrounding facilities. included. The vehicle position detection may be performed by using an autonomous navigation sensor such as a gyro sensor or a vehicle speed sensor in combination with the GPS 12.

The TV tuner processing unit 14 receives broadcast signals such as terrestrial digital broadcasting and performs processing for reproducing video and audio. The radio tuner processing unit 16 receives a radio broadcast signal and performs a process of reproducing audio. The AV processing unit 18 performs a process of reading and playing back music data and video data that have been compressed and stored in the hard disk device 70. As the music data and the video data, those read from a CD or DVD using a disc reading device (not shown) may be used, or those received via a network may be used.

The operation unit 20 is for receiving various operations by the user, and is provided with various switches, operation knobs, and the like. The input control unit 22 monitors the operation state of the operation unit 20 and detects the input content by the user.

The display processing unit 24 outputs video signals for displaying various operation screens, input screens, etc., displays these screens on the display device 26, and displays the video screens corresponding to the broadcast signals received by the TV tuner processing unit 14. A video signal for displaying a video screen or the like reproduced by the AV processing unit 18 is output, and these screens are displayed on the display device 26. The display device 26 is installed in front of the center of the driver's seat and the passenger seat, and is configured by using, for example, a liquid crystal display (LCD).

The microphone 30 collects the uttered voice of the user (for example, the driver of the own vehicle). The analog-digital converter 32 converts the audio signal collected by the microphone 30 into digital utterance data.

The digital-to-analog converter 40 converts the guidance sound and audio sound (digital data) generated by the processing of the navigation processing unit 10, the TV tuner processing unit 14, the radio tuner processing unit 16, and the AV processing unit 18 into analog sound. It is converted into a signal and output from the speaker 42. In reality, an amplifier that amplifies the signal is connected between the digital-to-analog converter 40 and the speaker 42, but this amplifier is omitted in FIG. Further, although the combinations of the digital-to-analog converter 40 and the speaker 42 are provided for the number of reproduction channels, only one set is shown in FIG.

The control unit 50 is for controlling the entire in-vehicle device 1, and is realized by executing a predetermined program stored in a ROM, RAM, or the like on the CPU. The control unit 50 includes a menu creation unit 51, a voice recognition processing unit 52, a recognition result selection unit 53, a viewing information acquisition unit 54, a viewing history creation unit 55, a viewing channel priority setting unit 56, and a facility usage information acquisition unit 57. It has a facility usage history creation unit 58 and a facility priority setting unit 59.

The menu creation unit 51 creates a menu screen including various commands (a plurality of words to be voice recognition) to be voice recognition. In addition, various menu screens necessary for the operation of the navigation processing unit 10 and the like are created. This menu screen is hierarchized, and when one word included in the first displayed menu screen is selected, the next menu screen corresponding to this word is displayed.

The voice recognition processing unit 52 extracts features from the voice uttered by the user collected by the microphone 30, and a plurality of recognition candidates registered in the voice recognition dictionary stored in the hard disk device 70. By comparing with each of the words, a voice recognition process is performed to select a word having a higher degree of similarity than a predetermined threshold value. In the present embodiment, the voice uttered by the user is constantly collected, and the voice recognition process is performed on the voice from the beginning of the collected voice to the time when the utterance is completed.

For example, considering the case where the user utters "Ichanneru" (1 channel) and recognizes this content by performing voice recognition processing, when the first "I" occurs, the voice "I" is heard. Since there is a large difference from the recognition word "1 channel", the voice recognition process will fail. Similarly, even when "Icha", "Icchan", "Icchanne", etc. are uttered, the voice recognition process fails because there is a large difference between these and the recognition word "1 channel". Then, when "Icchanneru" is uttered, the difference between the recognition word "1 channel" to be compared becomes small, so that the voice recognition process succeeds.

In order to perform such processing, the concepts of "distance" and "score" have been introduced. The distance indicates the similar distance between the voice as a comparison target and the recognition word. For example, as shown in FIG. 2, the maximum value of the distance is set to "1000" as the largest similarity distance. The score S indicates a similar distance between the actual input voice and the recognition word “1 channel”. As described above, for example, when the user utters "Icchanneru", the score at the start of utterance is 1000, and the score of the input voice gradually decreases as the utterance progresses. .. Then, when the score of the input voice at the time when the utterance is completed becomes less than the predetermined threshold value Th, the recognition result that the content of the input voice matches the recognition word is obtained. In general, the same processing is performed in parallel for each of a plurality of recognition words, and finally one recognition word is extracted as a recognition result. If the score of the input voice at the end of the utterance does not fall below the predetermined threshold value Th (when there is no recognition word close to the input voice), the voice recognition process has failed and the voice recognition process has failed again. Processing such as prompting voice input is performed.

By the way, considering the case where the user utters a voice having the same content as the recognition word stored in the voice recognition dictionary, the final score S should be a small value, and the threshold value Th is set to a certain degree. However, it should be possible to recognize the content of the voice uttered by the user. However, in reality, the final score S of the voice uttered by the user is quiet because the road noise, wind noise, rain sound, audio sound, etc. are collected at the same time as the voice uttered by the user. The value is not as small as when collecting sound in an environment. In consideration of this point, the threshold value Th for determining the success or failure of the voice recognition process cannot be set to a very small value. On the flip side, if the voice recognition process is performed using the threshold value Th set in this way, the difference between the threshold value Th and the score S is the case where the voice recognition process is successful. If is small, the environment around the user is loud with road noise, wind noise, rain noise, audio noise, etc., and if the difference between the threshold Th and the score S is large, the surroundings of the user are around the user. However, it can be said that the environment is quiet with little road noise, wind noise, rain noise, audio noise, etc.

The talk switch or the like may be operated by the user himself / herself to instruct the start and end of the utterance, and the voice recognition process may be performed with the input voice during that time as the recognition target.

When a plurality of words (scores less than a predetermined threshold value) determined to have high similarity are detected by the voice recognition processing unit 52, the recognition result selection unit 53 has the highest priority among them. Words are selected based on the user's hobbies and preferences and past history. This priority is set assuming the order in which the user is likely to select. In general, it is considered that the user's history reflects the user's hobbies and tastes. The word selection operation according to this priority may be performed when the difference in the similarity (score) of the detected plurality of words is within a predetermined value. Further, as this priority, the latest one at the time of word selection (at the time of collecting utterance voice using the microphone 30) is used.

The viewing information acquisition unit 54 acquires information (viewing information) of the television broadcast to be viewed using the TV tuner processing unit 14. This viewing information includes the channel number, genre, reception time, etc. of the received television broadcast. In addition, although it is desirable to acquire this viewing information for each user, considering that the viewing frequency of a specific user (for example, the driver of a vehicle) is often outstanding, this specific user It may be treated as viewing information corresponding to (it is not necessary to know who this user is).

The viewing history creation unit 55 creates a viewing history for each user by using the viewing information acquired by the viewing information acquisition unit 54. The created viewing history is stored in the hard disk device 70.

The viewing channel priority setting unit 56 sets the priority when the recognition result selection unit 53 selects a word related to the viewing channel based on the viewing history created by the viewing history creation unit 55.

The facility use information acquisition unit 57 acquires information on the facility (facility use information) used for the navigation operation of the navigation processing unit 10. This facility usage information includes information on facilities set as destinations for route search (store name, franchise chain name, genre, etc.), information on facilities actually used (parked) by the own vehicle (store name, etc.), etc. Is included.

The facility use history creation unit 58 creates a facility use history of the user by using the facility use information acquired by the facility use information acquisition unit 57. The created facility usage history is stored in the hard disk device 70.

The facility priority setting unit 59 sets the priority when the recognition result selection unit 53 selects words related to facility input during navigation operation based on the facility use history created by the facility use history creation unit 58. ..

The microphone 30 described above corresponds to the sound collecting means, the voice recognition processing unit 52 corresponds to the voice recognition means, and the recognition result selection unit 53 corresponds to the recognition result selection means.

The vehicle-mounted device 1 of the present embodiment has such a configuration, and the operation thereof will be described next. FIG. 3 is a flow chart showing an operation procedure for performing voice recognition processing in a state where menu screens corresponding to various commands to be voice recognition are displayed in the vehicle-mounted device 1.

When the menu screen created by the menu creation unit 51 is displayed on the display device 26 (step 100), the voice recognition processing unit 52 performs voice recognition processing on the user's uttered voice collected by the microphone 30. While doing so (step 102), it is determined whether or not the voice recognition is successful (step 104). When the score corresponding to the spoken voice is larger than the predetermined value (when the similarity with each word in the voice recognition dictionary is small), a negative determination is made, and the process returns to step 102 to continue the voice recognition process.

Further, when the score becomes less than a predetermined value (when the degree of similarity with each word in the speech recognition dictionary is large), an affirmative judgment is made in the determination in step 104. Next, the recognition result selection unit 53 determines whether or not a plurality of words have been detected by the voice recognition process (step 106). It is desirable that there is one word as a recognition result corresponding to the spoken voice, but if the readings of multiple commands to be recognized by voice included in the menu screen are similar, the score is threshold. Multiple words less than the value may be detected at the same time. In such a case, a positive judgment is made.

Next, the recognition result selection unit 53 determines whether or not the score difference between the detected plurality of words is within a predetermined value (step 108). The case where the score difference is within a predetermined value is a case where the scores of the words are close to each other, and there is a high possibility of erroneous recognition regardless of which one is selected. In this case, a positive judgment is made. Next, the recognition result selection unit 53 selects the word having the highest priority from the plurality of words as the final recognition result (step 110). After that, the process corresponding to this word is performed (step 112).

On the other hand, when there is only one word whose score is less than the predetermined value (negative judgment in the determination in step 106), or when the score difference between a plurality of words is larger than the predetermined value (negative judgment in the determination in step 108). The recognition result selection unit 53 selects the word with the highest score (step 114). After that, the process corresponding to this word is performed (step 112).

Next, a specific example of the priority for selecting one of a plurality of words when they are detected will be described.

FIG. 4 is a diagram showing the relationship between a channel designation command (word) and a priority of each channel when a desired television broadcast (channel) is designated by using the TV tuner processing unit 14.

When the user wants to select the channel designation command "1 channel", he / she can say "Icchanneru" by voice. However, when the user uttered "Icchanneru", the scores of the three words "1 channel", "8 channels", and "10 channels" whose voice data characteristics were relatively close were below the threshold value. These three words may be obtained as a recognition result. In such a case, "1 channel" having the highest priority is adopted as the final recognition result according to the priority set at that time.

By the way, the priority shown in FIG. 4 is set by the viewing channel priority setting unit 56 based on the viewing history created by the viewing history creating unit 55 and stored in the hard disk device 70. As an example, the following four cases can be considered.

(Case A1) Increase the priority of channels with high viewing frequency (long cumulative viewing time). For example, the viewing frequency for the most recent month is extracted based on the viewing history, and this priority is set.

(Case A2) Set a high priority for TV broadcast channels of genres with high viewing frequency (long cumulative viewing time). Information that identifies the genre of the program being broadcast at that time can be obtained from the broadcast signal being received, and the viewing channel priority setting unit 56 has three words "1 channel" and "8 channels". , The priority of each channel is set by comparing the genre of each channel corresponding to "10 channels" with the reception time of each genre included in the viewing history.

(Case A3) The priority of the TV broadcast channel currently being viewed is set low. When switching channels is assumed, it is desirable to lower the priority because the probability of specifying the channel being received is low.

(Case A4) When external information such as disaster information that is not directly related to the content of the TV broadcast being received is received, the priority of the channel that is broadcasting the program of the genre in which this external information is easily available is set high. To do. For example, when disaster information is received, the priority of the channel currently broadcasting the news program is set high.

In addition, these four cases may be combined appropriately. For example, a case A1 and a case A3 may be combined, or a case A2 and a case A3 may be combined.

FIG. 5 is a diagram showing an example of a search result screen in which a nearby convenience store is extracted using the navigation processing unit 10. On the search result screen, convenience stores existing around the vehicle position are displayed in order of proximity to the vehicle position, including the store name (franchise chain name).

FIG. 6 is a diagram showing the relationship between the designated command (word) when designating the convenience store to be visited on the search result screen shown in FIG. 5 and the priority of each word.

When the user wants to select the facility "AA Mart" displayed at the top of the search result screen, he / she can say "Go to the first" by voice. However, when the user says "go to the first", the four words "go to the first", "go to the second", "go to the third", and "go to the third" have relatively similar characteristics of the voice data. In some cases, the score for "I'm going to mess up" is below the threshold, and these four words are obtained as recognition results. In such a case, "Go to Mitsume", which has the highest priority, is adopted as the final recognition result according to the priority set at that time.

By the way, the priority shown in FIG. 6 is set by the facility priority setting unit 59 based on the facility use history created by the facility use history creation unit 58 and stored in the hard disk device 70. As an example, the following four cases can be considered.

(Case B1) Raise the priority of convenience stores belonging to franchise chains that are frequently used (frequently used). For example, the frequency of use in the last month is extracted based on the facility usage history, and this priority is set.

(Case B2) Set a high priority for convenience stores that are close to the vehicle position. Since the location of convenience stores is often known around homes, etc., when searching for convenience stores, it is assumed that the vehicle is traveling outside the vicinity of homes, etc. For this reason, it is conceivable that they are trying to finish their errands by searching for a nearby convenience store for the time being. In addition, these two cases may be combined appropriately.

As described above, in the in-vehicle device 1 of the present embodiment, when a plurality of words are obtained as the voice recognition result, among them according to the priority set assuming the order in which the user is likely to select. By selecting one from, it is possible to reduce erroneous recognition in speech recognition.

In addition, if the difference in similarity between a plurality of words obtained by the voice recognition process is small, there is a high possibility that erroneous recognition will occur. False recognition in voice recognition can be reduced.

In addition, by setting the priority based on the hobbies and tastes of the user who speaks and the past history, it is possible to set the priority of the word in consideration of the possibility of the user speaking.

In addition, by setting the priority based on the user's situation when performing voice recognition processing, the user can speak according to the user's operation content and surrounding environment during voice recognition processing. It is possible to set the priority of words in consideration of gender.

In addition, it is desirable to set the priority at the time when the spoken voice to be voice recognition is collected. By using the latest priority, it is possible to further reduce erroneous recognition in speech recognition.

The present invention is not limited to the above embodiment, and various modifications can be made within the scope of the gist of the present invention. For example, in the above-described embodiment, specific examples are shown for a case of designating a channel during reception of a television broadcast (FIG. 4) and a case of designating one of the convenience stores searched in the navigation operation (FIGS. 5 and 6). However, these are examples, and the present invention can be applied to others.

For example, the operation of synthesizing the images captured by the cameras provided on the front, rear, left and right sides of the vehicle to generate and display the top view image seen from above, and the back view image of the rear image captured by the rear camera behind the vehicle are generated. When any of the actions to be displayed is specified by voice, "top view" is uttered as a word for selecting the former, and "back view" is uttered as a word for selecting the latter. When the user utters "○○ view" and the scores of the two words "top view" and "back view" are both below the threshold value, the priority setting unit (not shown) is the steering angle. These priorities may be set based on (snake angle). For example, if the steering angle is large, it is possible that you are trying to enter an empty space in the parking lot or your garage while turning the steering wheel, so the priority of "back view" is higher than that of "top view". Set. In addition, when the steering angle is small, after such an approaching motion, the vehicle may be retracted while checking the distance between the left and right and rear walls and other vehicles. Set the priority of "view" higher than that of "back view".

Further, in the above-described embodiment, a plurality of words whose scores are less than the threshold value are extracted by the voice recognition process, and when the score difference between them is equal to or less than a predetermined value, selection is performed according to the priority. When multiple words with scores below the threshold value are extracted, there are many ambient noises such as road noise, wind noise, rain noise, and audio sound, and each score is close to the threshold value. It is thought that it occurs in. Therefore, the recognition result selection unit 53 may determine the level of ambient noise and perform word selection operation when the noise level exceeds the reference value.

As described above, according to the present invention, when a plurality of words are obtained as a voice recognition result, one of them is set according to the priority set assuming the order in which the user is likely to select. By selecting one, it is possible to reduce erroneous recognition in voice recognition.

1 In-vehicle device 10 Navigation processing unit 14 TV tuner processing unit 16 Radio tuner processing unit 18 AV processing unit 30 Microphone 32 Analog-to-digital converter (A / D)
51 Menu creation unit 52 Voice recognition processing unit 53 Recognition result selection unit 54 Viewing information acquisition unit 55 Viewing history creation unit 56 Viewing channel priority setting unit 57 Facility usage information acquisition unit 58 Facility usage history creation unit 59 Facility priority setting unit 70 Hard disk device (HDD)

Claims (10)

A sound collecting means for collecting the user's spoken voice,
A voice recognition dictionary in which multiple words to be selected are registered, and
A voice that extracts features from the spoken voice collected by the sound collecting means and selects a word having a higher degree of similarity than a predetermined threshold value from the plurality of words registered in the voice recognition dictionary. Recognition means and
When there are a plurality of the words determined to have high similarity by the voice recognition means, the recognition result selection means for selecting the word having the highest priority among them, and the recognition result selection means.
A voice recognition device characterized by comprising. The recognition result selection means is a case where there are a plurality of the words determined to have a high degree of similarity by the voice recognition means, and the difference in the degree of similarity between the plurality of words is within a predetermined value. The voice recognition device according to claim 1, wherein the word having the highest priority is selected from the above. The voice recognition device according to claim 1 or 2, wherein the priority is set based on the hobbies and tastes of the user who speaks. The voice recognition device according to any one of claims 1 to 3, wherein the priority is set based on a past history of the voice recognition result. The voice recognition device according to any one of claims 1 to 4, wherein the priority is set based on a situation of a user when performing voice recognition processing. The voice recognition device according to any one of claims 1 to 5, wherein the priority is set at the time when the spoken voice to be voice recognition is collected by the sound collecting means. Features are extracted from the spoken voice collected by the voice recognition means, and a word having a higher similarity than a predetermined threshold is selected from a plurality of words registered in the voice recognition dictionary by the voice recognition means. Voice recognition steps and
When there are a plurality of the words determined to have high similarity by the voice recognition means, the recognition result selection step of selecting the word having the highest priority from the words by the recognition result selection means.
A voice recognition processing method characterized by having. The recognition result selection means is a case where there are a plurality of the words determined to have a high degree of similarity by the voice recognition means, and the difference in the degree of similarity between the plurality of words is within a predetermined value. The voice recognition processing method according to claim 7, wherein the word having the highest priority is selected from the above. Computer,
A voice recognition means that extracts features from the spoken voice collected by the sound collecting means and selects a word having a higher degree of similarity than a predetermined threshold value from the plurality of words registered in the voice recognition dictionary. When,
When there are a plurality of the words determined to have high similarity by the voice recognition means, the recognition result selection means for selecting the word having the highest priority among them, and the recognition result selection means.
A voice recognition processing program that works.
The recognition result selection means is a case where there are a plurality of the words determined to have a high degree of similarity by the voice recognition means, and the difference in the degree of similarity between the plurality of words is within a predetermined value. The voice recognition processing program according to claim 9, wherein the word having the highest priority is selected from the above.

Images