JP6817915B2 - Speech recognition devices, in-vehicle systems and computer programs - Google Patents

Description

The present invention relates to a voice recognition technique for recognizing a user's spoken voice.

As a voice recognition technology for recognizing a user's voice, a first voice that recognizes a user's voice picked up by a microphone in a system equipped with an audio source device that outputs the voice of audio content such as music from a speaker. In addition to the recognition unit, a second voice recognition unit that performs voice recognition of the voice output by the audio source device to the speaker is provided, and the result of voice recognition by the first voice recognition unit and the voice recognition by the second voice recognition unit. There is known a technique for invalidating the result of voice recognition by the first voice recognition unit when the results match the results (for example, Patent Document 1).

According to such a technique, it is possible to prevent the voice recognition result of the first voice recognition unit recognized for the output voice of the audio source device that wraps around the microphone as the voice recognition result of the user's spoken voice. can do.

Jitsuto 2602342

By the way, when performing voice recognition of the voice picked up by the microphone, the score indicating the magnitude of the difference between the voice picked up by the microphone and the voice of the recognition candidate is used as a guide for a plurality of recognition candidates, which are words and phrases, from the microphone. In parallel with the voice input, the voice score of each recognition candidate for the input voice is sequentially calculated in real time, and the calculated score for the voice of any recognition candidate is a predetermined threshold value. When it becomes the following, it is conceivable to calculate the recognition candidate whose score is below the threshold as the recognition result.

However, in the case of performing such voice recognition, if the above-mentioned technique of providing the first voice recognition unit and the second voice recognition unit is applied, the following problems occur.
That is, in this case, when the audio source device outputs a voice of the same phrase as any of the recognition candidates, the voice is picked up by the microphone together with other voice components such as noise, so that the first voice recognition The score with the recognition candidate calculated by the unit is generally larger than the score with the recognition candidate calculated by the second voice recognition unit for the same voice output by the audio source device.

As a result, the calculation time of the recognition result of the first voice recognition unit for the output voice of the same phrase as the recognition candidate of the audio source device is delayed from the calculation time of the recognition result of the second voice recognition unit for the voice. It will end up.

Therefore, in this case, in order to invalidate the voice recognition result of the first voice recognition unit recognized for the output voice of the audio source device that wraps around the microphone, the recognition result of the second voice recognition unit is used. The period from the calculation time to the elapse of the above-mentioned delay time is set as the adjustment period, and for the result of voice recognition by the first voice recognition unit during the adjustment period, the first voice recognition unit makes a voice. It is investigated whether or not there is a match between the recognized result and the result of voice recognition by the second voice recognition unit prior to the voice recognition of the first voice recognition unit, and if they match, the first voice recognition unit It is necessary to perform a process to invalidate the result of voice recognition.

On the other hand, the delay time described above varies depending on the content and environment of the output audio of the audio source device and is unspecified, so that the time length of the adjustment period described above cannot be uniquely determined. If the time length of the adjustment period here is too short, the voice recognition result of the first voice recognition unit recognized for the output voice of the audio source device cannot be invalidated, and the time length of the adjustment period is long. If is too long, even the voice recognition result of the voice actually spoken by the user will be invalidated.

Therefore, an object of the present invention is to perform voice recognition that recognizes only the voice spoken by the user more accurately in an environment where the sound of the audio source device is radiated from the speaker.

In order to achieve the above object, the present invention is arranged in the space in a voice recognition device that recognizes the voice uttered in the space in which the sound output from the audio source device is emitted from the speaker. The microphone, the first voice recognition means for inputting the voice picked up by the microphone, and voice recognition for recognizing a phrase that is expected to match the voice in parallel with the input of the voice, and the voice. A second voice recognition means that inputs the voice output to the speaker by the source device and recognizes words and phrases that are expected to match the voice output to the speaker in parallel with the input of the voice. , The first speech recognition means is provided with a recognition adjusting means for outputting a phrase recognized as a recognition result. Here, if the second voice recognition means recognizes the phrase, the second voice recognition means detects the completion of the output of the voice of the recognized word from the audio source, and the recognition adjustment means is the second voice recognition means. If the word is recognized, the recognized word is set as the adjustment word, and thereafter, until the second voice recognition means detects the completion of the output, or the second voice recognition means completes the output. The output as the recognition result of the same phrase as the adjusted phrase recognized by the first speech recognition means is suppressed until a predetermined period elapses after the detection of.

Further, in order to achieve the above-mentioned problem, the present invention provides a voice recognition device for recognizing a sound uttered in a space in which a sound output from an audio source device is radiated from the speaker. The microphone arranged in the above and the voice picked up by the microphone are input, and each time the sound of each voice section input from the microphone is input for a plurality of recognition candidates which are words and phrases, the evaluation value of the recognition candidate is calculated. , When the sound of the voice section is consistent with the sound of the section corresponding to the voice section of the voice that pronounced the recognition candidate, it is decreased, and when it is not consistent, it is increased and the evaluation value is With respect to the plurality of recognition candidates by inputting the first voice recognition means for performing voice recognition for recognizing the words and phrases of the recognition candidates having become equal to or less than a predetermined first threshold and the sound output to the speaker by the audio source device. Each time the sound of each voice section input from the audio source device is input, the evaluation value of the recognition candidate is set to the section corresponding to the voice section of the voice in which the sound of the voice section sounds the recognition candidate. It decreases when it matches the sound, increases it when it does not match, recognizes the words and phrases of the recognition candidate whose evaluation value is equal to or less than the predetermined second threshold value, and recognizes the words and phrases of the recognition candidate. A second voice recognition means for detecting the occurrence of a peak in which the evaluation value changes from a decrease to an increase after recognition, and a recognition adjustment means for outputting a phrase recognized by the first voice recognition means as a recognition result. Is. Here, if the second voice recognition means recognizes a word, the recognition adjusting means sets the recognized word as the adjusting word, and thereafter, the second voice recognition means detects the occurrence of the peak. Or, until a predetermined period of time elapses after the second voice recognition means detects the occurrence of the peak, the output as the recognition result of the same word as the adjustment word recognized by the first voice recognition means is suppressed. To do.

Here, in such a voice recognition device, it is also preferable to set a value larger than the first threshold value as the second threshold value.
Further, in order to achieve the above-mentioned problem, the present invention provides a voice recognition device for recognizing a sound uttered in a space in which a sound output from an audio source device is radiated from the speaker. The microphone arranged in the above and the voice picked up by the microphone are input, and each time the sound of each voice section input from the microphone is input for a plurality of recognition candidates which are words and phrases, the evaluation value of the recognition candidate is calculated. , The sound of the voice section is increased when the sound of the voice that sounds the recognition candidate is consistent with the sound of the section corresponding to the voice section, and is decreased when the sound is not matched, and the evaluation value is With respect to the plurality of recognition candidates by inputting the first voice recognition means for performing voice recognition for recognizing the words and phrases of the recognition candidate having reached a predetermined first threshold value or more and the sound output by the audio source device to the speaker. Each time the sound of each voice section input from the audio source device is input, the evaluation value of the recognition candidate is set to the section corresponding to the voice section of the voice in which the sound of the voice section sounds the recognition candidate. It increases when it matches the sound, decreases it when it does not match, recognizes the words and phrases of the recognition candidate whose evaluation value is equal to or higher than the predetermined second threshold value, and recognizes the words and phrases of the recognition candidate. A device including a second voice recognition means for detecting the occurrence of a peak in which the evaluation value changes from an increase to a decrease after recognition, and a recognition adjustment means for outputting a phrase recognized by the first voice recognition means as a recognition result. Is. Here, if the second voice recognition means recognizes a word, the recognition adjusting means sets the recognized word as the adjusting word, and thereafter, the second voice recognition means detects the occurrence of the peak. Or, until a predetermined period of time elapses after the second voice recognition means detects the occurrence of the peak, the output as the recognition result of the same word as the adjustment word recognized by the first voice recognition means is suppressed. Here, in such a voice recognition device, it is also preferable to set a value smaller than the first threshold value as the second threshold value.

The present invention also provides an in-vehicle system including the above voice recognition device, the speaker mounted on the automobile, and the audio source device. Here, in this in-vehicle system, the space becomes the vehicle interior space of the automobile.

In the above-mentioned voice recognition system or in-vehicle system, if the second voice recognition means recognizes a phrase, the completion of output of the voice of the phrase recognized by the second voice recognition means from the audio source is subsequently detected. Or, until a predetermined period of time elapses after detecting the completion of the output, the output as the recognition result of the same phrase recognized by the first speech recognition means and recognized by the second speech recognition means is suppressed. To.

Here, the timing at which the first voice recognition means recognizes a phrase of the voice output by the audio source device output from the speaker is during the period during which the audio source device outputs the voice of the phrase. Further, the first voice recognition means is a voice output by the audio source device from a voice in which the voice output from the audio source device output from the microphone and other voices such as noise are mixed. Since the phrase is recognized, the timing at which the first voice recognition means recognizes the phrase is the timing after the second voice recognition means recognizes the phrase.

Therefore, according to the above-mentioned voice recognition system or in-vehicle system, the second voice recognition means recognizes only during the period in which the first voice recognition means can recognize the words and phrases of the voice output by the audio source device. Since it is possible to suppress the output of the recognition result of the same phrase as the phrase recognized by the voice recognition means, it is possible to suppress the output of the phrase recognized from the voice output from the speaker by the audio source device as the recognition result of the user's spoken voice. At the same time, it becomes possible to correctly output the words and phrases recognized from the voice actually spoken by the user as the recognition result of the voice spoken by the user.

As described above, according to the present invention, in an environment where the sound of the audio source device is radiated from the speaker, it is possible to perform voice recognition that recognizes only the voice spoken by the user more correctly.

It is a block diagram which shows the structure of the information processing system which concerns on embodiment of this invention. It is a figure which shows the method of voice recognition which concerns on embodiment of this invention. It is a flowchart which shows the recognition adjustment process which concerns on embodiment of this invention. It is a figure which shows the processing example of the recognition adjustment processing which concerns on embodiment of this invention.

Hereinafter, an embodiment of the voice recognition device according to the embodiment of the present invention will be described by taking as an example an application to an information processing system mounted on an automobile.
FIG. 1 shows the configuration of the information processing system according to the present embodiment. As shown in the figure, the information processing system includes a microphone 1, a first voice recognition engine 2, a second voice recognition engine 3, a voice recognition dictionary 4, and a recognition adjustment unit. 5. It includes a voice input control unit 6, one or more applications 7 such as a navigation application, an audio source 8, and a speaker 9.

In such a configuration, the audio source 8 is a device that serves as a sound source for a radio receiver, a music player, or the like, and outputs the sound of the audio content to the speaker 9 and the second voice recognition engine 3.

Then, the speaker 9 radiates the sound input from the audio source 8 into the vehicle.
Further, the voice recognition dictionary 4 is a dictionary for voice recognition, and a plurality of recognition candidates, which are different words and phrases, and pronunciation data representing the pronunciation are registered. The pronunciation data may represent a phoneme sequence of pronunciation of a phrase, or may be voice data of pronunciation of a phrase or the like.

Next, the first voice recognition engine 2 uses the voice recognition dictionary 4 to perform voice recognition processing on the voice input from the microphone 1 and outputs the voice-recognized words to the recognition adjustment unit 5 as recognition words.

Further, the second voice recognition engine 3 uses the voice recognition dictionary 4 to perform voice recognition processing on the voice input from the audio source 8, and outputs the words and phrases recognized by the voice to the recognition adjustment unit 5 as recognition words.

Then, the recognition adjustment unit 5 calculates the words and phrases to be the recognition result using the recognition words and phrases input from the first voice recognition engine 2 and the recognition words and phrases input from the second voice recognition engine 3, and the calculated recognition result is voiced. Output to the input control unit 6. Here, the operation of the recognition adjustment unit 5 will be described in detail later.

Then, the voice input control unit 6 outputs the voice input corresponding to the recognition result output by the recognition adjustment unit 5 to the application 7, and the application 7 accepts the voice input and processes according to the content of the received voice input. I do.

Hereinafter, the operation of voice recognition performed by the first voice recognition engine 2 and the second voice recognition engine 3 will be described.
The first voice recognition engine 2 and the second voice recognition engine 3 are recognized voices (voice output by the microphone 1 if the first voice recognition engine 2 is used, and voice output by the audio source 8 if the second voice recognition engine 3 is used. ), The score of each recognition candidate stored in the speech recognition dictionary 4 for the speech to be recognized is calculated.

Here, the recognition candidate score for the recognition target voice represents a predicted value of the magnitude of the difference between the recognition target voice and the utterance data of the recognition candidate, and the score is higher when a larger difference is predicted. Becomes larger.

More specifically, in the calculation of the score, after setting a predetermined initial value as the score, each time the sound of each voice section of the recognition target voice (for example, the voice section for each phoneme) is input, The presence or absence of matching between the sound of the voice section and the portion of the pronunciation data of each recognition candidate corresponding to the voice section is calculated, and if they match, the score is reduced by a predetermined value, and if they do not match, the score is determined. This is done by increasing the value.

According to such voice recognition, in FIG. 2a, when the recognition target voice is "aiueoka", the transition of the score calculated for the recognition candidate "aiueo" and the recognition candidate "aiueoka" As shown by the transition of the score calculated by the above, while the sound of the recognition target voice that matches the recognition candidate is input, the score with the recognition candidate gradually decreases, and the sound of the recognition target voice that does not match the recognition candidate. While is entered, the recognition candidate scores are gradually increased.

That is, for example, as shown in FIG. 2a1, the recognition target voice "aiueoka", the recognition candidate "aiueo", and the score gradually decrease during the period in which the recognition target voice "aiueo" sound is input. After that, it increases when the "ka" of the recognition target voice is input.

Similarly, as shown in FIG. 2a2, the recognition target voice "aiueoka", the recognition candidate "aiai", and the score gradually decrease during the period in which the recognition target voice "ai" sound is input. After that, the period during which the recognition target voice "Eoka" is input gradually increases.

By the way, in the first voice recognition engine 2 and the second voice recognition engine 3, if the score of the recognition target voice calculated as described above and one of the recognition candidates becomes the threshold value Th or less, A recognition candidate word whose score is equal to or less than the threshold value Th is recognized and output to the recognition adjustment unit 5 as a recognition word.

That is, for example, in the case shown in FIG. 2a1, the score for the recognition candidate "aiueo" is equal to or less than the threshold value Th immediately before the "e" of the recognition target voice "aiueoka" is input. Then, the recognition candidate "aiueo" is output to the recognition adjustment unit 5 as a recognition phrase.

On the other hand, in the case shown in FIG. 2a1, since the score for the recognition candidate "Au Aio" does not fall below the threshold value Th, the phrase of the recognition candidate "Ai Aio" is not recognized.

In the above voice recognition, even if the score between the recognition target voice and one of the recognition candidates is equal to or less than the threshold value Th, the difference between the recognition candidate's score and the recognition candidate is smaller than a predetermined level. If there are other recognition candidates that have been calculated, recognition is not performed at that point, and the difference between the recognition candidate for which the minimum score is calculated and the scores of the other recognition candidates becomes larger than a predetermined level. At that time, the recognition candidate word for which the small score is calculated is recognized and output to the recognition adjustment unit 5 as the recognition word, or the end of the voice input to the microphone 1 is waited for at that point. The phrase of the recognition candidate for which the minimum score is calculated may be recognized and output to the recognition adjustment unit 5 as the recognition phrase.

Next, FIG. 2b shows an operation example of voice recognition of the first voice recognition engine 2 and the second voice recognition engine 3 performed on the output voice of the audio source 8.
As shown in the figure, when the audio source 8 outputs the output voice of "What is it", this output voice is output from the speaker 9, picked up by the microphone 1, and the first voice recognition is performed for this output voice. Voice recognition is performed in the engine 2.

On the other hand, the output voice of the audio source 8 of this "Nanunune no ha" is directly sent to the second voice recognition engine 3, and the second voice recognition engine 3 also performs voice recognition.

Then, FIG. 2b1 is calculated for the recognition candidate “Nanune no” in the voice recognition of the first voice recognition engine 2 for the output voice of “Nani Nune no ha” by such an audio source 8. 2b2 shows the transition of the score, and FIG. 2b2 shows the same recognition candidate “Nannu” in the voice recognition of the second voice recognition engine 3 for the output voice of “Nanune no ha” by the audio source 8. It shows the transition of the score calculated for "Neno".

In this case, as shown in the figure, the first speech recognition engine 2 is the recognition candidate "Nanunune" during the period when "Nanunenno" of the output voice "Nanunenano" of the audio source 8 is input. The score calculated for "no" and the score calculated by the second speech recognition engine 3 for the recognition candidate "what" will gradually decrease, but it will be calculated by the first speech recognition engine 2. The degree of decrease in the score is smaller than the degree of decrease in the score calculated by the second speech recognition engine 3. This is because the voice input to the first voice recognition engine 2 is the output voice of the audio source 8 picked up by the microphone 1 together with other voice components such as noise, so that each voice section of the voice output from the microphone 1 The consistency of the pronunciation data of the recognition candidate "Nani Nuneno" with the part corresponding to the relevant voice section is the recognition candidate "reception candidate" for each voice section of the output voice of the audio source 8 that is directly input to the second voice recognition engine 3. This is because it is smaller than the consistency of the pronunciation data of "What is it" with the part corresponding to the voice section.

Further, in the present embodiment, as the threshold value Th set in the first speech recognition engine 2, a value Th1 smaller than the threshold value Th2 set in the first speech recognition engine 2 is set.
Therefore, as shown in FIGS. 2b1 and 2b2, the score calculated for the recognition candidate "Nanune" in the second speech recognition engine 3 is the recognition candidate "Nanune" in the first speech recognition engine 2. It becomes the threshold value Th or less at a time earlier than the score calculated for ". Therefore, after the recognition candidate "what" is recognized by the second speech recognition engine 3 and output to the recognition adjustment unit 5 as a recognition phrase, the recognition candidate "na" is delayed by the first speech recognition engine 2. "Ninune no" is recognized and output to the recognition adjustment unit 5 as a recognition phrase.

In the present embodiment, the threshold value Th1 set in the first speech recognition engine 2 is set to a value Th1 smaller than the threshold value Th2 set in the first speech recognition engine 2. This is to ensure that the voice recognition for the output voice of the audio source 8 is performed in the second voice recognition engine 3 before the first voice recognition engine 2.

By the way, if the second speech recognition engine 3 recognizes the recognition candidate whose score is equal to or less than the threshold value Th2 and outputs it as the recognition phrase to the recognition adjustment unit 5, the recognition candidate as the recognition phrase is calculated. The subsequent transition of the score is monitored, and as shown in FIG. 2b2, if a downward peak (a point at which the score changes from decreasing to increasing) appears in the waveform of the transition of the score, this is detected and the recognition adjustment unit A process of notifying 5 of the detection of the peak is also performed.

The operation of voice recognition performed by the first voice recognition engine 2 and the second voice recognition engine 3 has been described above.
Hereinafter, using the recognition phrase input from the first speech recognition engine 2 and the recognition phrase input from the second speech recognition engine 3 performed by the recognition adjustment unit 5 as described above, the phrase to be the recognition result is calculated and calculated. The operation of outputting the recognized recognition result to the voice input control unit 6 will be described.

FIG. 3 shows a procedure of the recognition adjustment process performed by the recognition adjustment unit 5.
As shown in the figure, in this process, the recognition adjustment unit 5 generates a recognition phrase input from the first speech recognition engine 2 (step 302) and a recognition phrase input from the second speech recognition engine 3 (step 302). Step 304) and the occurrence of input of the peak detection notification from the second voice recognition engine 3 (step 306) are monitored.

Then, if the recognition phrase input from the second speech recognition engine 3 occurs (step 304), the mask period is set (step 312), and the recognition phrase input from the second speech recognition engine 3 is used as the adjustment phrase. Set (step 314). Then, the process returns to the monitoring of steps 302, 304, and 306.

On the other hand, when the input of the peak detection notification from the second voice recognition engine 3 occurs (step 306), the setting during the mask period is cleared (step 308), and the adjustment phrase setting is cleared (step 310). .. Then, the process returns to the monitoring of steps 302, 304, and 306.

Further, when the recognition phrase is input from the first speech recognition engine 2 (step 302), it is checked whether the mask period is set (step 316), and if the mask period is not set, it is checked. , The recognition phrase input from the first speech recognition engine 2 is calculated as a phrase to be the recognition result, and the calculated recognition result is output to the speech input control unit 6 (step 320). Then, the process returns to the monitoring of steps 302, 304, and 306.

On the other hand, in step 316, when it is determined that the mask period is set, it is checked whether the recognition phrase input from the first speech recognition engine 2 matches the adjustment phrase (step 318), and one If so, the recognition phrase input from the first speech recognition engine 2 is discarded, and the process returns to the monitoring of steps 302, 304, and 306 as it is.

On the other hand, when the recognition phrase input from the first speech recognition engine 2 does not match the adjustment phrase (step 318), the recognition phrase input from the first speech recognition engine 2 is calculated as the phrase to be the recognition result. , The calculated recognition result is output to the voice input control unit 6 (step 320). Then, the process returns to the monitoring of steps 302, 304, and 306.

The recognition adjustment process performed by the recognition adjustment unit 5 has been described above.
Next, a processing example of such recognition adjustment processing is shown in FIG.
FIG. 4 shows recognition of the recognition candidates “Jitaku ni Kaeru” of the first voice recognition engine 2 and the second voice recognition engine 3 performed on the output voice “Jitaku ni Kaeru Hito ...” of the audio source 8. It shows the operation.

When the audio source 8 outputs the output voice of "Human who changes ...", this output voice is output from the speaker 9, picked up by the microphone 1, and the first voice recognition engine 2 responds to this output voice. , As shown in FIG. 4a, the score for the recognition candidate "Jitaku ni Kaeru" is calculated.

Further, when the audio source 8 outputs the output voice of "Human who changes to human ...", this output voice is directly sent to the second voice recognition engine 3, and also in the second voice recognition engine 3, FIG. 4b. As shown in, the score for the recognition candidate "Jitaku ni Kaeru" is calculated.

In this case, during the period in which the audio source 8 inputs the output voice "Jitaku ni Kaeru" of "Jitaku ni Kaeru", the first speech recognition engine 2 responds to the knowledge candidate "Jitaku ni Kaeru". The score to be calculated and the score calculated by the second speech recognition engine 3 for the recognition candidate "Jitaku ni Kaeru" gradually decrease, but the degree of decrease calculated by the first speech recognition engine 2 is the first. 2 It becomes smaller than the degree of decrease calculated by the voice recognition engine 3.

Further, the first speech recognition engine 2 is set as the threshold value Th1 which is smaller than the threshold value Th2 set in the first speech recognition engine 2.
Therefore, the score calculated for the recognition candidate "Jitaku ni Kaeru" in the second speech recognition engine 3 is higher than the score calculated for the recognition candidate "Jitaku ni Kaeru" in the first speech recognition engine 2. At an early point, the threshold value becomes Th or less, and at time t21, the second speech recognition engine 3 recognizes the recognition candidate "Jitaku ni Kaeru" and outputs it to the recognition adjustment unit 5 as a recognition phrase.

Then, when the recognition phrase "jitaku ni kaeru" is output from the second speech recognition engine 3 at time t21, this recognition phrase "jitaku ni kaeru" is set as the adjustment phrase and the mask period is set. To.

After that, the second speech recognition engine 3 monitors the transition of the score calculated for the recognition candidate "Jitaku ni Kaeru", and at time t22, the downward peak of the score transition waveform (from the decrease in the score). When a point that turns to increase) appears, the second voice recognition engine 3 detects the peak and notifies the recognition adjustment unit 5 of the detection of the peak.

Then, when the recognition adjustment unit 5 is notified of the detection of the peak, the recognition adjustment unit 5 clears the setting during the mask period.
On the other hand, at time t21, the recognition candidate "jitaku ni kaeru" is recognized by the second speech recognition engine 3 and output to the recognition adjustment unit 5 as a recognition phrase, and then the recognition candidate "jitaku ni kaeru" is also recognized by the first speech recognition engine 2. The score calculated for "Takuni Kaeru" becomes equal to or less than the threshold value Th, and at time t11, the recognition candidate "Jitakuni Kaeru" is recognized by the first speech recognition engine 2 and output to the recognition adjustment unit 5 as a recognition phrase. Will be done.

Here, at the time t11 when the recognition candidate "jitaku ni kaeru" is recognized by the first voice recognition engine 2 and output to the recognition adjustment unit 5 as a recognition phrase, the voice "t11" is the voice output by the audio source 8. It will occur during the period when the part that matches the recognition candidate "Jitaku ni Kaeru" is being output.

On the other hand, at the time t22 when the second voice recognition engine 3 detects the peak, the audio source 8 outputs the voice "Jitaku ni Kaeru Hito ...", which is a recognition candidate "Jitaku ni Kaeru Hito ...". It is the end of the period when the part that matches "" is output.

Therefore, the time point t11 when the recognition candidate "jitaku ni kaeru" is recognized by the first voice recognition engine 2 and output to the recognition adjustment unit 5 as a recognition phrase is a time point during the period in which the mask period is set. ..

By the way, when the recognition phrase "jitaku ni kaeru" is output from the first voice recognition engine 2 at the time point t11, the recognition adjustment unit 5 is currently set during the mask period, so that the first voice recognition engine Compare the recognition phrase "Jitaku ni Kaeru" output from 2 with the set adjustment phrase "Jitaku ni Kaeru". Then, in this case, the recognition phrase "jitaku ni kaeru" output from the first voice recognition engine 2 and the set adjustment phrase "jitaku ni kaeru" match, so that the first voice recognition engine 2 Discard the recognition phrase "Jitaku ni Kaeru" output from, without making it a recognition result.

As a result, the recognition result of the user's uttered voice of the recognition phrase "Jitaku ni Kaeru" recognized by the first voice recognition engine 2 with respect to the voice "Jitaku ni Kaeru Hito ..." output by the audio source 8. The output to the voice input control unit 6 is suppressed.

The embodiment of the present invention has been described above.
In the above embodiment, in the recognition adjustment process of the recognition adjustment unit 5, the setting during the mask period is canceled when the second voice recognition engine 3 notifies the detection of the peak. As shown in FIG. 4b, the setting during the mask period may be canceled at the time t23 when the predetermined margin time mgn elapses from the time t22 when the detection of the peak is notified from the second voice recognition engine 3.

By doing so, it is possible to more reliably suppress the output to the voice input control unit 6 as the recognition result of the recognition phrase recognized by the first voice recognition engine with respect to the voice output by the audio source 8. Will be.

Further, in the above embodiment, in the recognition adjustment process of the recognition adjustment unit 5, the setting during the mask period is canceled when the second voice recognition engine 3 notifies the detection of the peak, but during the mask period If the setting is canceled during the mask period, the voice that matches the voice of the recognition phrase output by the second voice recognition engine 3 is output when the audio source 8 ends. , Any other method may be used. That is, for example, based on the number of phonemes output from the audio source 8, the output of the voice matching the voice of the recognition phrase is detected at the time when the audio source 8 ends, and the setting during the mask period is canceled. May be good. ,

Also in this case, the setting during the mask period is set when the predetermined margin time mgn elapses from the time when the audio source 8 ends the output of the voice matching the voice of the recognition phrase output by the second voice recognition engine 3. You may try to release it.

Further, the above embodiment may be carried out by reversing the positive and negative directions of the score.
That is, the score of the recognition candidate for the recognition target voice may be calculated so as to be smaller as the difference is predicted to be larger.
More specifically, in the calculation of the score, after setting a predetermined initial value as the score, each time the sound of each voice section of the recognition target voice (for example, the voice section for each phoneme) is input, The presence or absence of matching between the sound of the voice section and the portion of the pronunciation data of each recognition candidate corresponding to the voice section is calculated, and if they match, the score is increased by a predetermined value, and if they do not match, the score is determined. The value may be reduced.

However, in this case, in the first voice recognition engine 2 and the second voice recognition engine 3, the scores of the recognition target voice calculated as described above and one of the recognition candidates are equal to or higher than the threshold value Th. If so, the recognition candidate word / phrase whose score is equal to or higher than the threshold value Th is recognized and output to the recognition adjustment unit 5 as a recognition phrase. Further, the threshold value Th2 set as the second voice recognition engine 3 threshold value Th is set to a value smaller than the threshold value Th1 set as the first voice recognition engine 2 threshold value Th. Further, if the second speech recognition engine 3 recognizes the recognition candidate whose score is equal to or less than the threshold value Th2 and outputs it as a recognition phrase to the recognition adjustment unit 5, the recognition candidate as the recognition phrase is calculated. The subsequent transition of the score is monitored, and if an upward peak (a point at which the score changes from an increase to a decrease) appears in the waveform of the transition of the score, the peak is detected and the recognition adjustment unit 5 is notified of the detection of the peak. ..

The voice recognition technology in the above embodiments can be applied not only to an information processing system mounted on an automobile but also to an arbitrary information processing system that performs voice input.

1 ... Microphone, 2 ... 1st voice recognition engine, 3 ... 2nd voice recognition engine, 4 ... Voice recognition dictionary, 5 ... Recognition adjustment unit, 6 ... Voice input control unit, 7 ... Application, 8 ... Audio source, 9 ... Speaker.

Claims (7)

It is a voice recognition device that recognizes the voice uttered in the space where the sound output from the audio source device is emitted from the speaker to the speaker.
With the microphone placed in the space,
A first voice recognition means for inputting a voice picked up by the microphone and performing voice recognition for recognizing a phrase that is predicted to match the voice in parallel with the input of the voice.
A second voice recognition that inputs the voice output to the speaker by the audio source device and recognizes a phrase that is predicted to match the voice output to the speaker in parallel with the input of the voice. Means and
It has a recognition adjusting means that outputs a phrase recognized by the first voice recognition means as a recognition result.
When the second voice recognition means recognizes the phrase, the second voice recognition means detects the completion of the output of the voice of the recognized phrase from the audio source, and detects the completion of the output.
If the second voice recognition means recognizes a word, the recognition adjusting means sets the recognized word as the adjusting word, and thereafter, until the second voice recognition means detects the completion of the output, or , The output as the recognition result of the same phrase as the adjusted phrase recognized by the first speech recognition means is suppressed until a predetermined period elapses after the second speech recognition means detects the completion of the output. A featured voice recognition device. It is a voice recognition device that recognizes the voice uttered in the space where the sound output from the audio source device is emitted from the speaker to the speaker.
With the microphone placed in the space,
Each time the voice picked up by the microphone is input and the sound of each voice section input from the microphone is input for a plurality of recognition candidates that are words and phrases, the evaluation value of the recognition candidate is evaluated by the sound of the voice section. , Decrease when the voice that sounds the recognition candidate matches the sound of the section corresponding to the voice section, increase when it does not match, and the evaluation value is the predetermined first threshold value. The first voice recognition means for recognizing the following words and phrases of recognition candidates, and
Each time the voice output from the audio source device is input to the speaker and the sound of each voice section input from the audio source device is input for the plurality of recognition candidates, the evaluation value of the recognition candidate is set to the voice section. When the sound of is consistent with the sound of the section corresponding to the voice section of the voice that pronounced the recognition candidate, it is decreased, and when it is not consistent, it is increased, and the evaluation value is a predetermined second. A second speech recognition means that recognizes the words and phrases of the recognition candidate that are below the threshold value and detects the occurrence of a peak in which the evaluation value changes from decrease to increase after recognizing the words and phrases of the recognition candidate.
It has a recognition adjusting means that outputs a phrase recognized by the first voice recognition means as a recognition result.
If the second voice recognition means recognizes a word, the recognition adjusting means sets the recognized word as the adjusting word, and thereafter, until the second voice recognition means detects the occurrence of the peak, or , The output of the same phrase as the adjusted phrase recognized by the first speech recognition means as a recognition result is suppressed until a predetermined period of time elapses after the second speech recognition means detects the occurrence of the peak. A featured voice recognition device. The voice recognition device according to claim 2.
A voice recognition device characterized in that a value larger than the first threshold value is set as the second threshold value. It is a voice recognition device that recognizes the voice uttered in the space where the sound output from the audio source device is emitted from the speaker to the speaker.
With the microphone placed in the space,
Each time the voice picked up by the microphone is input and the sound of each voice section input from the microphone is input for a plurality of recognition candidates that are words and phrases, the evaluation value of the recognition candidate is evaluated by the sound of the voice section. , The voice that pronounces the recognition candidate is increased when it is consistent with the sound of the section corresponding to the voice section, and is decreased when it is not consistent, and the evaluation value is a predetermined first threshold value. The first voice recognition means for recognizing the words and phrases of the recognition candidates described above, and
Each time the voice output from the audio source device is input to the speaker and the sound of each voice section input from the audio source device is input for the plurality of recognition candidates, the evaluation value of the recognition candidate is set to the voice section. When the sound of is consistent with the sound of the section corresponding to the voice section of the voice that pronounced the recognition candidate, it is increased, and when it is not consistent, it is decreased, and the evaluation value is a predetermined second. A second voice recognition means for recognizing the words and phrases of the recognition candidate that have exceeded the threshold value and detecting the occurrence of a peak in which the evaluation value changes from an increase to a decrease after recognizing the words and phrases of the recognition candidate.
It has a recognition adjusting means that outputs a phrase recognized by the first voice recognition means as a recognition result.
If the second voice recognition means recognizes a word, the recognition adjusting means sets the recognized word as the adjusting word, and thereafter, until the second voice recognition means detects the occurrence of the peak, or , The output of the same phrase as the adjusted phrase recognized by the first speech recognition means as a recognition result is suppressed until a predetermined period of time elapses after the second speech recognition means detects the occurrence of the peak. A featured voice recognition device. The voice recognition device according to claim 4.
A voice recognition device characterized in that a value smaller than the first threshold value is set as the second threshold value. The voice recognition device according to claim 1, 2, 3, 4 or 5 mounted on an automobile.
It has the speaker mounted on the automobile and the audio source device.
An in-vehicle system characterized in that the space is an interior space of the automobile. A computer program that is read and executed by a computer equipped with a microphone arranged in a space where sound output from an audio source device is radiated from the speaker.
The computer program uses the computer,
A first voice recognition means for inputting a voice picked up by the microphone and performing voice recognition for recognizing a phrase that is predicted to match the voice in parallel with the input of the voice.
Second voice recognition that inputs the voice output to the speaker by the audio source device and recognizes words and phrases that are predicted to match the voice output to the speaker in parallel with the input of the voice. Means and
The phrase recognized by the first speech recognition means is made to function as a recognition adjustment means for outputting as a recognition result.
If the second voice recognition means recognizes the phrase, the second voice recognition means detects the completion of the output of the voice of the recognized phrase from the audio source, and detects the completion of the output.
If the second voice recognition means recognizes a word, the recognition adjusting means sets the recognized word as the adjustment word, and thereafter, until the second voice recognition means detects the completion of the output, or The second speech recognition means suppresses the output as a recognition result of the same phrase recognized by the first speech recognition means until a predetermined period of time elapses after detecting the completion of the output. Computer program.