JP6759058B2 - Voice recognition device and voice recognition method - Google Patents

Description

The present invention relates to a voice recognition device and a voice recognition method, and in particular, calculates an index indicating the degree of similarity between the voice pattern of a word registered in a voice recognition dictionary and the spoken voice input from a microphone, and the calculation is performed. It is suitable for use in a voice recognition device that recognizes a word corresponding to an uttered voice depending on whether or not the index satisfies a predetermined condition with respect to the threshold value.

The vehicle is equipped with various electronic devices such as audio equipment, air conditioners, and navigation devices. Further, in order to avoid one-handed driving when operating these electronic devices, a system is also provided in which the electronic devices can be operated by voice recognition. By using this voice recognition technology, the driver can operate various electronic devices without taking his / her hand off the steering wheel (without manually operating the operation unit such as the remote controller or the operation panel).

The voice recognition device usually recognizes a specific word, a compound word, a simple command sentence, etc. (hereinafter, collectively referred to as "word") uttered by the user as a utterance command. The electronic device controls according to the word (speech command) recognized by the voice recognition device. In such a voice recognition device, an acoustic model in which each recognition target word used as an utterance command and its voice pattern are associated is registered in advance in a voice recognition dictionary. Then, the feature amount calculated from the user's spoken voice is compared with the feature amount of the acoustic model to search for the voice pattern having the highest similarity, and the word associated with the voice pattern is the spoken voice word. Recognize that.

In the conventional voice recognition device, the user presses the built-in utterance button to enter the voice recognition mode, recognizes the user's utterance voice input from the microphone, and executes a command. It is also known that instead of operating the utterance button, a specific action such as clapping a hand triggers the voice recognition mode. Recently, a voice recognition device (hereinafter referred to as a triggerless voice recognition device) that does not require a trigger such as an operation of a utterance button or a specific operation during voice recognition has been provided.

In the triggerless voice recognition device, the microphone is always on, the input voice is identified, and it is determined whether or not the word corresponds to the utterance command. That is, the distance value is calculated as an index indicating the degree of closeness (similarity) between the voice pattern of each word registered in the voice recognition dictionary and the voice input from the microphone. Then, when the calculated distance value is smaller than the threshold value set for each word, the input voice is recognized as a word below the threshold value.

In the passenger compartment, the voice input from the microphone includes various noises such as engine operation sound, running sound, audio voice, and conversation voice between passengers, in addition to the spoken voice for voice recognition. ing. In particular, in the case of a triggerless voice recognition device, a voice recognition mode is not provided, and noise that becomes noise is always input to the microphone. Therefore, it is necessary to devise a way to correctly perform voice recognition even in such an environment.

In order to increase the correct answer rate of speech recognition (suppress false recognition), it is necessary to appropriately set the threshold value to be compared with the distance value. On the other hand, conventionally, the noise level in the vehicle interior is estimated based on the driving parameters of the vehicle (engine speed, vehicle speed, strength of the blower fan of the in-vehicle air conditioner, output volume of the car stereo, etc.), and the estimated noise level. A voice recognition device is known that sets a threshold value for voice recognition according to the above (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2001-75595

However, in the technique described in Patent Document 1, the threshold value is variably set based on the noise level, and the content of noise is not considered. That is, even if the noise level is the same, the correct answer rate of speech recognition varies depending on the content of the noise, but since this is not taken into consideration in the technique of Patent Document 1, there is a problem that the threshold value cannot be optimized. there were.

The present invention has been made to solve such a problem, and it is possible to more appropriately set a threshold value to be compared with an index of speech recognition, so that the occurrence of false recognition can be reduced. The purpose is to.

In order to solve the above-mentioned problems, in the present invention, an index indicating the degree of similarity between the voice pattern of the word registered in the voice recognition dictionary and the spoken voice input from the microphone is calculated, and the calculated index is used. The source type of the audio voice being played in the vehicle is determined and determined in the voice recognition device that recognizes the word corresponding to the spoken voice depending on whether or not a predetermined condition is satisfied with respect to the threshold value. The threshold is variably set according to the source type.

According to the present invention configured as described above, noise is set because a threshold value to be compared with the voice recognition index is set according to the source type of the audio voice that becomes noise when the word to be recognized is voice-recognized. The threshold value can be optimized according to the content of, and the occurrence of erroneous recognition can be reduced.

It is a functional block diagram which shows the structural example of the voice recognition apparatus by 1st Embodiment. It is a figure which shows an example of the table information which a threshold value setting part refers to. It is a flowchart which shows the operation example of the voice recognition apparatus by 1st Embodiment. It is a flowchart which shows the operation example of the voice recognition apparatus by 1st Embodiment. It is a functional block diagram which shows the structural example of the voice recognition apparatus by 2nd Embodiment. It is a flowchart which shows the operation example of the voice recognition apparatus by 2nd Embodiment. It is a functional block diagram which shows the structural example of the voice recognition apparatus by 3rd Embodiment. It is a flowchart which shows the operation example of the voice recognition apparatus by 3rd Embodiment. It is a figure which shows another example of the table information which a threshold value setting part refers to.

(First Embodiment)
Hereinafter, the first embodiment according to the present invention will be described with reference to the drawings. FIG. 1 is a functional block diagram showing a configuration example of the voice recognition device 100 according to the first embodiment. The voice recognition device 100 of the present embodiment recognizes the user's utterance voice (words such as a specific word, a compound word, and a simple command sentence) input from the microphone 200 as a utterance command, and the utterance command to the navigation device 300. Is to execute. Although the electronic device to be controlled is the navigation device 300 here, it may be an audio device 400, an air conditioner, or other electronic device.

As shown in FIG. 1, the voice recognition device 100 according to the first embodiment includes a recognition dictionary storage unit 11, a voice recognition unit 12, a confirmation unit 13, a source determination unit 14, and a threshold value setting unit 15. .. Each of the above functional blocks 12 to 15 can be configured by any of hardware, DSP (Digital Signal Processor), and software. For example, when configured by software, each of the above functional blocks 12 to 15 is actually configured to include a computer CPU, RAM, ROM, etc., and is a program stored in a recording medium such as RAM, ROM, hard disk, or semiconductor memory. Is realized by the operation of.

The recognition dictionary storage unit 11 associates a word to be recognized with its voice pattern, and stores a voice recognition dictionary in which a threshold value to be compared with a voice recognition index is set. In the present embodiment, as an example of the index, a distance value (for example, a degree of closeness) indicating the degree of closeness (similarity) between the voice pattern of each word registered in the recognition dictionary storage unit 11 and the spoken voice input from the microphone 200 (for example). , A value from 0 to 1000) is used. The smaller the distance value, the higher the similarity. In the present embodiment, the threshold value is variably set.

The voice recognition unit 12 calculates an index indicating the degree of similarity between the voice pattern of the word registered in the voice recognition dictionary of the recognition dictionary storage unit 11 and the utterance voice input from the microphone 200, and the calculated index is used. When a predetermined condition is satisfied with respect to the threshold value, the spoken voice is recognized as a word satisfying the predetermined condition. As described above, when the distance value is used as an index, the voice recognition unit 12 sets the threshold value for the spoken voice when the distance value calculated for the spoken voice is smaller than the threshold value set for each word. Recognize that the word is lower. When an index whose value increases as the degree of similarity increases is used, the voice recognition unit 12 uses the spoken voice when the index calculated for the spoken voice is larger than the threshold value set for each word. Is recognized as a word that exceeds the threshold value.

By presenting the word recognized by the voice recognition unit 12 to the user, the confirmation unit 13 causes the user to confirm whether or not the word matches the word uttered by the user. This confirmation is performed, for example, by a talkback that outputs a word recognized by the voice recognition unit 12 as a synthetic voice. Alternatively, the word recognized by the voice recognition unit 12 may be displayed on the screen in characters.

If a word different from the word uttered by the user is presented by the confirmation unit 13, the user can instruct to cancel and cancel the voice recognition. The cancellation instruction can be given by manual operation of the touch panel or utterance input of the word "cancel". If the user does not instruct the cancellation within a predetermined time after the voice recognition result is presented by the confirmation unit 13, the confirmation unit 13 determines the word recognized by the voice recognition unit 12 and sends the navigation device 300 as an utterance command. Output.

The source determination unit 14 determines the source type of the audio sound reproduced by the audio device 400 in the vehicle. Audio-voice is a voice that becomes noise for voice recognition. Here, among audio-visual sounds, those that are likely to contain many conversations (for example, voices such as news and dramas) are less likely to contain many conversations (for example, music-type voices). There are various sources up to. It can be said that the more conversations are included, the higher the possibility that the voice recognition unit 12 will cause erroneous recognition.

Therefore, the source determination unit 14 classifies the source types according to the high possibility that many conversations are included, and determines which of the classified source types the audio / voice being reproduced in the vehicle corresponds to. judge. The source type of the audio audio reproduced by the audio device 400 can be determined by checking the source setting information of the audio device 400.

The threshold value setting unit 15 variably sets the voice recognition threshold value stored in the recognition dictionary storage unit 11 according to the source type determined by the source determination unit 14. FIG. 2 is a diagram showing an example of table information in which the classification of the source type and the adjustment value of the threshold value for each classification are associated with each other. The threshold value setting unit 15 variably sets the threshold value for voice recognition with reference to this table information.

In the example of FIG. 2, the source types are classified into three according to the high possibility that many conversations are included. The first category is a music source such as a USB (Universal Serial Bus) to which a CD (Compact Disc), a memory card, a portable sound source, or the like is connected. The second category is a video source such as DVD (Digital Versatile Disk), HDMI (High-Definition Multimedia Interface), and AUX. The third category is news / drama sources such as DTV (Digital TeleVision) and Radio.

Here, in the order of the first division <second division <third division, there is a high possibility that many conversations are included in the audio voice reproduced by the audio device 400. In the present embodiment, the table information is set so as to lower the threshold value of voice recognition as the category is more likely to contain conversations. As the threshold value becomes smaller, the distance value calculated by the voice recognition unit 12 is less likely to fall below the threshold value, so that the occurrence of erroneous recognition can be reduced.

The numerical value shown in FIG. 2 indicates a value to be adjusted with respect to the reference threshold value. That is, in the case of a music source, it is shown that "40" is added to the reference threshold value. In the case of a video source, it is shown that the reference threshold value is used as it is. In the case of news / drama sources, it indicates that "20" is subtracted from the reference threshold.

3 and 4 are flowcharts showing an operation example of the voice recognition device 100 according to the first embodiment. The flowchart shown in FIG. 3 starts when the power of the voice recognition device 100 is turned on, and is continuously executed until the power of the voice recognition device 100 is turned off. The flowchart shown in FIG. 4 shows a specific processing content in step S2 of FIG. Here, an operation example of triggerless voice recognition in which the microphone 200 is always on and the voice recognition unit 12 always performs voice recognition even if the user does not perform any particular operation is shown.

In FIG. 3, first, the voice recognition unit 12 and the confirmation unit 13 perform voice recognition processing (step S1). That is, the voice recognition unit 12 calculates a distance value indicating the degree of similarity between the voice pattern of the word registered in the voice recognition dictionary of the recognition dictionary storage unit 11 and the spoken voice input from the microphone 200, and the calculation is performed. Recognize words whose distance value is less than the threshold. Then, the confirmation unit 13 presents the word recognized by the voice recognition unit 12 to the user, and if there is no cancellation instruction within a predetermined time, the confirmation unit 13 outputs the recognized word as an utterance command to the navigation device 300.

Next, the source determination unit 14 and the threshold value setting unit 15 execute the threshold value setting process according to the source type (step S2). That is, in FIG. 4, the source determination unit 14 determines whether or not the audio audio is being reproduced in the audio device 400 (step S11). When the audio / audio is not reproduced, the processing of the flowchart shown in FIG. 4 ends.

When the audio audio is being reproduced, the source determination unit 14 determines the source type of the reproduced audio audio (step S12). Then, the threshold value setting unit 15 variably sets the voice recognition threshold value stored in the recognition dictionary storage unit 11 with reference to the table information shown in FIG. 2 according to the source type determined by the source determination unit 14. (Step S13). As a result, the processing of the flowchart shown in FIG. 4 is completed.

As described in detail above, in the first embodiment, the source type of the audio sound being reproduced in the vehicle is determined, and the voice recognition threshold value is variably set according to the determined source type. According to the first embodiment configured in this way, the similarity between the voice pattern of the registered word and the spoken voice is shown according to the source type of the audio voice that becomes noise when the word to be recognized is voice-recognized. Since the threshold value to be compared with the distance value is set, the threshold value can be optimized according to the content of noise, and the occurrence of erroneous recognition can be reduced.

(Second Embodiment)
Next, a second embodiment according to the present invention will be described with reference to the drawings. FIG. 5 is a functional block diagram showing a configuration example of the voice recognition device 100A according to the second embodiment. Note that, in FIG. 5, those having the same reference numerals as those shown in FIG. 1 have the same functions, and therefore, duplicate description will be omitted here.

As shown in FIG. 5, the voice recognition device 100A according to the second embodiment further includes a cancellation number counting unit 16. Further, the voice recognition device 100A according to the second embodiment includes a threshold value setting unit 15A instead of the threshold value setting unit 15.

The cancellation count unit 16 is used by the user within a predetermined time after the confirmation unit 13 presents the word recognized by the voice recognition unit 12 (the word whose distance value calculated for the spoken voice is smaller than the threshold value) to the user. The number of times the cancellation is instructed (hereinafter referred to as the number of cancellations) is counted. The cancellation count unit 16 stores the cancellation count for each word.

The threshold value setting unit 15A has the following functions in addition to the functions of the threshold value setting unit 15 described in the first embodiment. That is, when the number of cancellations counted by the number of cancellations counting unit 16 reaches a predetermined number, the threshold value setting unit 15A changes the threshold value in a direction that makes it difficult for the index calculated for the spoken voice to satisfy the predetermined condition.

Here, when a distance value is used as an index of similarity, the threshold value setting unit 15A changes the threshold value stored in the recognition dictionary storage unit 11 to a value smaller than the current value for words whose number of cancellations has reached a predetermined number. To do. For example, the threshold value setting unit 15A sets a value obtained by subtracting a predetermined value from the current value of the threshold value as a new threshold value. When an index is used in which the value increases as the degree of similarity increases, the threshold setting unit 15A sets the threshold value stored in the recognition dictionary storage unit 11 for words that have reached a predetermined number of cancellations from the current state. Is also changed to a value larger by a predetermined value.

When the user (for example, the driver) instructs to cancel, the word of the utterance command is included in the audio voice or the conversation voice of another passenger even though the driver has not uttered the word corresponding to the utterance command. It is recognized by the voice recognition unit 12, and the confirmation operation by the confirmation unit 13 has occurred. Moreover, the fact that such a situation occurs repeatedly a predetermined number of times means that the same situation may occur repeatedly in the future. In that case, the driver has to instruct the cancellation each time, which is troublesome. Therefore, in the second embodiment, the voice recognition unit 12 makes it difficult for the voice recognition unit 12 to recognize the word when the number of cancellations reaches a predetermined number by reducing the threshold value.

FIG. 6 is a flowchart showing an operation example of the voice recognition device 100A according to the second embodiment. FIG. 6 shows a specific processing example of step S1 in FIG.

First, the voice recognition unit 12 calculates a distance value indicating the degree of similarity between the voice pattern of the word registered in the voice recognition dictionary of the recognition dictionary storage unit 11 and the voice input from the microphone 200, and the calculation is performed. A word recognition process in which the distance value is smaller than the threshold value is executed (step S21). Then, the confirmation unit 13 determines whether or not a word whose distance value is smaller than the threshold value is detected by the voice recognition unit 12 in the voice recognition dictionary (step S22).

Here, when it is determined that the voice recognition unit 12 has not detected a word whose distance value is smaller than the threshold value, the processing of the flowchart shown in FIG. 6 ends, and the process proceeds to step S2 shown in FIG. On the other hand, when it is determined that the voice recognition unit 12 has detected a word whose distance value is smaller than the threshold value, the confirmation unit 13 presents the detected word and confirms with the user whether or not it matches the spoken voice. (Step S23).

After that, the confirmation unit 13 determines whether or not the user has instructed to cancel within a predetermined time (step S24). When the cancellation is instructed within the predetermined time, the cancellation count counting unit 16 increments the cancellation count value (step S25). Then, the threshold value setting unit 15A determines whether or not the number of cancellations has reached a predetermined number (step S26).

Here, when the number of cancellations reaches a predetermined number, the threshold setting unit 15A sets the threshold value stored in the recognition dictionary storage unit 11 smaller than the current value for the word for which the number of cancellations has reached the predetermined number. Change to a value (step S27). As a result, the processing of the flowchart shown in FIG. 6 is completed, and the process proceeds to the processing of step S2 shown in FIG. On the other hand, if the number of cancellations has not yet reached the predetermined number, the processing of the flowchart shown in FIG. 6 is completed without changing the threshold value, and the process proceeds to step S2 shown in FIG.

If it is determined in step S24 that there is no cancellation instruction within a predetermined time, the cancellation count counting unit 16 clears the cancellation count value to zero (step S28). As a result, the processing of the flowchart shown in FIG. 6 is completed, and the process proceeds to the processing of step S2 shown in FIG. In this case, the confirmation unit 13 outputs the recognized word as an utterance command to the navigation device 300.

As described in detail above, according to the second embodiment, when the user's cancel operation for the word recognized by the voice recognition unit 12 is repeated a predetermined number of times, the voice is reduced by reducing the threshold value for that word. It is made difficult for the recognition unit 12 to recognize the word. As a result, the situation in which the voice recognition unit 12 recognizes the word against the intention of the user is reduced, and the user does not have to perform the troublesome cancel operation many times.

(Third Embodiment)
Next, a third embodiment according to the present invention will be described with reference to the drawings. FIG. 7 is a functional block diagram showing a configuration example of the voice recognition device 100B according to the third embodiment. In addition, in FIG. 7, those having the same reference numerals as those shown in FIG. 5 have the same functions, and therefore, duplicate description will be omitted here.

As shown in FIG. 7, the voice recognition device 100B according to the third embodiment further includes a proximity count unit 17. Further, the voice recognition device 100B according to the third embodiment includes a threshold value setting unit 15B instead of the threshold value setting unit 15A.

The proximity count counting unit 17 counts the number of times (hereinafter, referred to as the proximity count) that the difference between the index and the threshold value is smaller than the predetermined value for words whose similarity index does not satisfy a predetermined condition with respect to the threshold value. Here, when a distance value is used as an index of similarity, the proximity count unit 17 determines the number of detections of words whose distance value is not smaller than the threshold value and whose difference between the distance value and the threshold value is smaller than a predetermined value. Count as the number of proximity. The proximity count unit 17 stores the proximity count for each word.

The fact that the difference between the distance value calculated for a certain word and the threshold value is smaller than the predetermined value means that the distance value does not fall below the threshold value and word recognition is not achieved, but the similarity to the registered word is relatively high. It means that the user is uttering a word in close proximity. For example, in the case where the user is uttering a registered word corresponding to the utterance command, but the distance value does not fall below the threshold value depending on the utterance state (volume, intonation, utterance speed, etc.) of the user, the distance value and the threshold value are used. The difference between and is smaller than the predetermined value.

The threshold value setting unit 15B has the following functions in addition to the functions of the threshold value setting unit 15A described in the second embodiment. That is, when the number of proximity counts by the proximity count unit 17 reaches a predetermined number of times, the threshold value setting unit 15B changes the threshold value in a direction in which the index calculated for the spoken voice easily satisfies the predetermined condition. As a result, in the third embodiment, the voice recognition unit 12 can easily recognize the word.

Here, when a distance value is used as an index of similarity, the threshold value setting unit 15B sets the threshold value stored in the recognition dictionary storage unit 11 by a predetermined value larger than the current value for words whose proximity count reaches a predetermined number of times. Change to a value. When an index is used in which the value increases as the degree of similarity increases, the threshold setting unit 15B sets the threshold value stored in the recognition dictionary storage unit 11 for words that have reached a predetermined number of proximity times from the current state. Is also changed to a smaller value by a predetermined value.

FIG. 8 is a flowchart showing an operation example of the voice recognition device 100B according to the third embodiment. FIG. 8 shows a specific processing example of step S1 in FIG. Note that, in FIG. 8, those having the same number as the step number shown in FIG. 6 perform the same processing, and therefore, duplicate description will be omitted here.

In step S22 of FIG. 8, when the confirmation unit 13 determines that a word whose distance value is smaller than the threshold value is detected in the voice recognition dictionary, the proximity count unit 17 clears the proximity count value to cello. (Step S29). After that, the process proceeds to step S23.

Further, in step S22, when the confirmation unit 13 determines that the word whose distance value is smaller than the threshold value is not detected in the voice recognition dictionary, the proximity count unit 17 increments the proximity count value. (Step S31). Then, the threshold value setting unit 15B determines whether or not the number of times of proximity has reached a predetermined number of times (step S32).

Here, when the number of proximity reaches a predetermined number of times, the threshold setting unit 15B increases the threshold value stored in the recognition dictionary storage unit 11 by a predetermined value for the word whose number of proximity reaches the predetermined number of times. Change to a value (step S33). As a result, the processing of the flowchart shown in FIG. 8 is completed, and the process proceeds to the processing of step S2 shown in FIG. On the other hand, when the number of proximity has not reached the predetermined number of times, the processing of the flowchart shown in FIG. 8 is completed without changing the threshold value, and the process proceeds to the process of step S2 shown in FIG.

As described in detail above, according to the third embodiment, when the distance value calculated for a certain word does not fall below the threshold value, but the difference from the threshold value is smaller than the predetermined value is repeated a predetermined number of times. By increasing the threshold value for the word, the voice recognition unit 12 can easily recognize the word. As a result, it is possible to eliminate the inconvenience that the user is uttering a word corresponding to the utterance command but is not recognized depending on the utterance state (volume, intonation, utterance speed, etc.) of the user.

In the first to third embodiments, an example in which the threshold value is variably set according to the source type of audio / audio has been described, but the present invention is not limited to this. For example, the threshold value may be variably set according to the combination of the audio-audio source type and the volume. FIG. 9 shows an example of table information to be referred to when the threshold value is variably set according to the combination of the source type and the volume. In the example of FIG. 9, the lower the volume, the more difficult it is to recognize the correct voice. Therefore, the table information is set so that the lower the volume, the larger the threshold value.

Further, in the third embodiment, the configuration in which both the cancellation count unit 16 and the proximity count unit 17 are provided has been described, but it is also possible to adopt an embodiment in which only the proximity count unit 17 is applied.

In addition, the first to third embodiments are merely examples of implementation of the present invention, and the technical scope of the present invention should not be interpreted in a limited manner by these. It must not be. That is, the present invention can be implemented in various forms without departing from its gist or its main features.

11 Recognition dictionary storage unit 12 Voice recognition unit 13 Confirmation unit 14 Source determination unit 15, 15A, 15B Threshold setting unit 16 Cancel count count unit 17 Proximity count count unit

Claims (5)

A recognition dictionary storage unit that stores a voice recognition dictionary that associates a word to be recognized with its voice pattern and sets a threshold value to be compared with a voice recognition index.
The above index indicating the similarity between the voice pattern of the word registered in the above voice recognition dictionary and the utterance voice input from the microphone is calculated, and the calculated index satisfies a predetermined condition with respect to the threshold. In this case, a voice recognition unit that recognizes that the uttered voice is a word that satisfies the above-mentioned predetermined conditions, and
A source judgment unit that determines the source type of audio audio being played in the vehicle,
A threshold setting unit that variably sets the threshold value according to the source type determined by the source determination unit ,
The number of cancellations, which is the number of times the user has instructed to cancel within a predetermined time after presenting to the user a word in which the index calculated for the spoken voice satisfies the predetermined condition with respect to the threshold value. Equipped with a counting section
The threshold value setting unit is characterized in that when the number of cancellations reaches a predetermined number, the threshold value is changed in a direction in which the index calculated for the spoken voice is less likely to satisfy the predetermined condition. Voice recognition device. A recognition dictionary storage unit that stores a voice recognition dictionary that associates a word to be recognized with its voice pattern and sets a threshold value to be compared with a voice recognition index.
The above index indicating the similarity between the voice pattern of the word registered in the above voice recognition dictionary and the utterance voice input from the microphone is calculated, and the calculated index satisfies a predetermined condition with respect to the threshold. In this case, a voice recognition unit that recognizes that the uttered voice is a word that satisfies the above-mentioned predetermined conditions, and
A source judgment unit that determines the source type of audio audio being played in the vehicle,
A threshold setting unit that variably sets the threshold value according to the source type determined by the source determination unit ,
For words whose index does not satisfy the predetermined condition with respect to the threshold value, a proximity count unit for counting the number of proximity times in which the difference between the index and the threshold value is smaller than the predetermined value is provided.
The threshold value setting unit is characterized in that when the number of times of proximity reaches a predetermined number of times, the threshold value is changed in a direction in which the index calculated for the spoken voice easily satisfies the predetermined condition. Voice recognition device. The source determination unit classifies the source types according to the high possibility that many conversations are included, and determines which of the classified source types the audio / voice being played in the vehicle corresponds to. The voice recognition device according to claim 1 or 2 , wherein the determination is made. The first step in which the source determination unit of the voice recognition device determines the source type of the audio voice being played in the vehicle,
The second step in which the threshold value setting unit of the voice recognition device variably sets the threshold value to be compared with the voice recognition index according to the source type determined by the source determination unit.
The voice recognition unit of the voice recognition device uses the threshold value set by the threshold value setting unit to determine the degree of similarity between the voice pattern of the word registered in the voice recognition dictionary and the spoken voice input from the microphone. A third step of calculating the index to be shown and recognizing that the spoken voice is a word satisfying the predetermined condition when the calculated index satisfies the predetermined condition with respect to the threshold value .
After the cancellation count unit of the voice recognition device presents to the user a word in which the index calculated for the spoken voice satisfies the predetermined condition with respect to the threshold value, the user instructs the user to cancel within a predetermined time. It has a fourth step of counting the number of cancellations, which is the number of times it has been done.
In the second step, when the number of cancellations reaches a predetermined number, the threshold setting unit changes the threshold in a direction in which the index calculated for the spoken voice is less likely to satisfy the predetermined condition. /> A voice recognition method characterized by this. The first step in which the source determination unit of the voice recognition device determines the source type of the audio voice being played in the vehicle,
The second step in which the threshold value setting unit of the voice recognition device variably sets the threshold value to be compared with the voice recognition index according to the source type determined by the source determination unit.
The voice recognition unit of the voice recognition device uses the threshold value set by the threshold value setting unit to determine the degree of similarity between the voice pattern of the word registered in the voice recognition dictionary and the spoken voice input from the microphone. A third step of calculating the index to be shown and recognizing that the spoken voice is a word satisfying the predetermined condition when the calculated index satisfies the predetermined condition with respect to the threshold value .
The proximity count unit of the voice recognition device counts the number of proximitys in which the difference between the index and the threshold is smaller than the predetermined value for words whose index does not satisfy the predetermined condition with respect to the threshold. Have a fourth step to
In the second step, when the number of times of proximity reaches a predetermined number of times, the threshold value setting unit changes the threshold value in a direction in which the index calculated for the spoken voice easily satisfies the predetermined condition. /> A voice recognition method characterized by this.