JPH01123299A - Voice dialing apparatus - Google Patents

Abstract

PURPOSE: To make the recognition rate higher for second speaking in the case of erroneous recognition for first speaking by removing an erroneously recognized word from the object of a dictionary in next speaking when a user indicates this erroneously recognized word in the recognition result and taking all words of the dictionary as the recognition object at the time of on-hook. CONSTITUTION: A dictionary mask part 8 masks only a specific word of a dictionary 6 as the recognition object to remove it from the recognition object. A receiver is hooked off, and first speaking is inputted and is recognized, and the recognition result is sent to a result control part 7. This part 7 not only informs the user of the result by display, voice, or the like but also prepares for automatic originating by an automatic originating part 10, and originating is successful in the case of correct recognition. The user depresses an erroneous recognition switch 9 in the case of erroneous recognition. In this case, originating is stopped, and the next candidate of the recognition result is shown. Several candidates are shown in this manner, and the user speaks again if they are all erroneously recognized. Thus, the name of a destination is correctly recognized with a high probability in next speaking though being erroneously recognized once.

Description

[Detailed description of the invention] Technical field The present invention relates to a voice dialing device.

Conventional voice dialing devices automatically call the other party's IT phone number based on the result of recognition from the voice uttered by the user, but the recognition rate is not 100% and there may be errors. give rise to recognition. Therefore, the recognition result is usually shown to the user using a display or audio output, and if there is a misrecognition, the user is notified using a switch or the like, and the automatic call is canceled. Once the user is misrecognized, he/she tries to make a call by saying the same name of the other party again. However, if you continue to utter the same utterances, the utterances will be more or less the same, and there is a high possibility that you will get the same erroneous recognition result.
It had the disadvantage of being extremely difficult to use for users.

1-1 Flag The present invention was made in view of the above-mentioned circumstances.
Especially in voice dialing equipment.

- This was done with the aim of increasing the recognition rate even higher the second time, if there is a misrecognition on the first utterance.

SUMMARY OF THE INVENTION In order to achieve the above object, the present invention includes a telephone receiver, a feature extraction section, a voice section detection section, and a recognition section.

The recognition results are used in a voice dialing device that has a result output control section, an erroneous recognition switch, a dictionary, a dictionary mask section, a means for notifying the user of the results, an on-hook/off-hook detection section, and an automatic transmission section. If the user indicates that the word has been misrecognized by the user, the word that was previously misrecognized will be removed from the dictionary in the next utterance, and once on-hook, the entire dictionary will be included in the recognition target. This is what I did.

Hereinafter, the present invention will be explained based on examples.

When using a voice dialing device, the user is considered to be speaking to make a call to the same destination from the time he or she picks up the handset until the time he or she hangs up the handset or dials the number.

Therefore, the recognition rate will be better if the words that have been notified of -degree misrecognition are excluded from subsequent recognition of utterances. Generally speaking, when they are uttered consecutively, the utterances are very similar, so
0, which is likely to result in the same false recognition result. For that reason,
- If words that have been misrecognized are removed from the dictionary, the next recognition rate will greatly increase. When the user wants to call another party, he or she can simply hang up the receiver, and of course, if the user automatically makes the - number of calls, there will be no words that are not included in the dictionary, and all words will be recognized.

FIG. 1 is a block diagram for explaining one embodiment of the present invention, and in the figure, 1 is a telephone receiver, and 2 is a hook.

3 is a feature extraction unit, 4 is a speech section detection unit, 5 is a recognition unit, 6
1 is a dictionary, 7 is a result control unit, 8 is a dictionary mask unit, 9 is an erroneous recognition switch, 10 is an automatic transmission unit, 11 is a display audio output unit, and the audio signal from the microphone of the receiver 1 is sent to a feature extraction unit 3. The data is input to the voice section detection section 4 and the respective data is sent to the recognition section 5. Further, the state of the hook 2 is sent from the receiver 1 to the dictionary mask section 8. The dictionary masking unit 8 can mask only specific words in the dictionary 6 to be recognized so as not to be recognized. When the handset is picked up, that is, when the handset goes off-hook, the dictionary mask section 8 first targets all words for recognition. Then, the -th utterance is input, recognition is performed, and the recognition result is sent to the result control section 7.

The result control section 7 notifies the user of the results through display or audio output, and the automatic transmission section 10 prepares for automatic transmission. If the recognition is correct, the transmission is successful.

In case of erroneous recognition, the user presses the erroneous recognition switch 9. In this case, the process is canceled and the next recognition result candidate is displayed. In this way, several candidates are shown, and if all of them are misrecognized, the utterance will be performed again. In this case, in the next recognition, the word that was previously misrecognized is made into a word that is not a recognition target in the dictionary masking unit 8. This is because if the user wants to call somewhere and has not done so already, he will naturally call the same number. Therefore, there is no need to use a word that is known to be incorrect as a recognition target again. In this way, recognition is performed by reducing the number of recognition targets, and the results are presented to the user. If a correct answer is given here, an automatic call is made and the dictionary mask section is cleared at the same time, allowing all words to be recognized next. However, if I misunderstood this time too, then...
When performing recognition, the next recognition is performed, excluding the erroneously recognized word this time, as in the previous time. If the user changes his or her mind and wants to place the call somewhere else, he or she can hang up the phone and pick it up again. Then, the hook goes from on-hook to off-hook, and the dictionary mask section is cleared by this hook signal.

Effects As is clear from the above explanation, according to the present invention, even if the name of the other party is misrecognized, it is possible to provide a voice dialing device that can correctly recognize the name of the other party with a higher probability in the next utterance.

[Brief explanation of the drawing]

FIG. 1 is a block diagram for explaining one embodiment of the present invention. DESCRIPTION OF SYMBOLS 1... Receiver, 2... Hook, 3... Feature extraction section, 4... Voice section detection section, 5... Recognition section, 6...
Dictionary, 7... Result control section, 8... Dictionary mask section, 9
... Misrecognition switch, 10... Automatic transmitter, 11.
...Display audio output section.

Claims (1)

[Claims] A telephone receiver, a feature extraction section, a voice section detection section, a recognition section, a result output control section, an erroneous recognition switch, a dictionary, a dictionary mask section, means for notifying the user of the results, and on-hook/off-hook detection. In a voice dialing device that has an automatic transmission section and an automatic transmission section, if the recognition result is indicated by the user as a misrecognition, the next time the user utters the word, the word that was previously indicated as an incorrect recognition is removed from the dictionary. A voice dialing device characterized in that when disconnected and on-hook, the entire dictionary is recognized.