Data processing: speech signal processing – linguistics – language – Speech signal processing – Recognition
Reexamination Certificate
1999-06-22
2001-10-23
Dorvil, Richemond (Department: 2741)
Data processing: speech signal processing, linguistics, language
Speech signal processing
Recognition
C704S240000, C704S255000, C704S275000
Reexamination Certificate
active
06308152
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a speech recognition method and a speech recognition apparatus for recognizing an uttered word and a speech control system for controlling an electric apparatus according to the recognized word.
2. Description of the Related Art
2.1. Previously Proposed Art
In a conventional speech recognition method, voice samples of a plurality of words desired to be recognized are registered as registered words in advance in a recognition word dictionary, and a word uttered by a user is recognized by using the recognition word dictionary. In this case, because it is difficult that a user knows all words registered in the recognition word dictionary, the user cannot avoid to utter a word other than the registered words. Therefore, even though the user utters a word other than the registered words, a specific word, of which an acoustic distance from the uttered word is shortest among those of the registered words, is selected from the registered words as a recognized word. As a result, in cases where the conventional speech recognition method is used for a conventional speech control system, there is a problem that an uttered word other than the registered words is erroneously recognized and an electric apparatus controlled by the speech control system is erroneously operated.
To prevent this problem, a word recognition score indicating a degree of an acoustic distance between the uttered word and a recognized word is calculated when the recognized word is determined, and the recognized word is adopted in cases where the word recognition score is higher than a threshold value. In contrast, in cases where the word recognition score is equal to or lower than the threshold value, the recognized word is rejected. That is, the recognized word is not adopted.
Therefore, an uttered word other than the registered words is not erroneously recognized because the word recognition score for the uttered word other than the registered words is low.
2.2. Problems to be Solved by the Invention
However, in cases where the word recognition score is calculated, it is required to adjust the threshold value according to environmental conditions (for example, noise conditions) of both the user and the speech control system. Also, it is required to set the threshold value changeable according to the combination of the registered words. Accordingly, there is a problem that it is difficult that the threshold value is set so as to reliably reject an uttered word differing from any registered words and to accurately recognize an uttered word agreeing with one of the registered words.
SUMMARY OF THE INVENTION
An object of the present invention is to provide, with due consideration to the drawbacks of such a conventional speech recognition method and a conventional speech control system, a speech recognition method and a speech recognition apparatus in which an uttered word differing from any registered words is reliably rejected and an uttered word agreeing with one registered word is accurately recognized as a recognized word even though a user does not know any registered words.
Also, an object of the present invention is to provide a speech control system in which an operation of an electric apparatus is correctly controlled according to the recognized word.
The object is achieved by the provision of a speech recognition method, comprising the steps of:
registering an acoustic feature of a recognition-desired word desired to be recognized for each of a plurality of recognition-desired words;
registering an acoustic feature of a reception word differing from the recognition-desired words for each of a plurality of recognition-desired words;
receiving an utterance including an uttered word;
calculating a recognition-desired word recognition score indicating a similarity degree between the uttered word and each recognition-desired word by comparing the acoustic feature of the recognition-desired word with an acoustic feature of the uttered word;
calculating a reception word recognition score indicating a similarity degree between the uttered word and each reception word by comparing the acoustic feature of the reception word with the acoustic feature of the uttered word;
recognizing the uttered word as a particular recognition-desired word corresponding to a particular recognition-desired word recognition score in cases where the particular recognition-desired word recognition score is higher than the highest reception word recognition score; and
rejecting the utterance in cases where the highest recognition-desired word recognition score is equal to or lower than the highest reception word recognition score.
Also, the object is achieved by the provision of a speech recognition apparatus, comprising:
recognition-desired word registering means for registering an acoustic feature of a recognition-desired word desired to be recognized for each of a plurality of recognition-desired words;
reception word registering means for registering an acoustic feature of a reception word differing from the recognition-desired words for each of a plurality of recognition-desired words;
word receiving means for receiving an utterance including an uttered word;
recognition-desired word recognition score calculating means for calculating a recognition-desired word recognition score indicating a similarity degree between the uttered word received by the word receiving means and each recognition-desired word registered by the recognition-desired word registering means by comparing the acoustic feature of the recognition-desired word with an acoustic feature of the uttered word;
reception word recognition score calculating means for calculating a reception word recognition score indicating a similarity degree between the uttered word received by the word receiving means and each reception word registered by the reception word registering means by comparing the acoustic feature of the reception word with the acoustic feature of the uttered word;
word recognizing means for recognizing the uttered word received by the word receiving means as a particular recognition-desired word corresponding to a particular recognition-desired word recognition score calculated by the recognition-desired word recognition score calculating means in cases where the particular recognition-desired word recognition score is higher than the highest reception word recognition score calculated by the reception word recognition score calculating means; and
utterance rejecting means for rejecting the utterance received by the word receiving means in cases where the highest recognition-desired word recognition score calculated by the recognition-desired word recognition score calculating means is equal to or lower than the highest reception word recognition score calculated by the reception word recognition score calculating means.
In the above steps and configuration, in cases where an utterance including an uttered word agrees with or is most similar to a particular recognition-desired word, a particular recognition-desired word recognition score corresponding to the particular recognition-desired word becomes highest among the recognition-desired word recognition scores and the reception word recognition scores. Therefore, the uttered word is recognized as the particular recognition-desired word.
In contrast, in cases where an uttered word included in an utterance is not most similar to any recognition-desired words but agrees with or is most similar to a particular reception word, a particular reception word recognition score corresponding to the particular reception word becomes highest among the recognition-desired word recognition scores and the reception word recognition scores. Therefore, the uttered word is rejected.
Accordingly, the uttered word can be reliably recognized at a high recognition efficiency.
Also, because it is not required to set a threshold value changeable with environmental conditions for the word recognition score, the uttered word can be easily recognized.
Konuma Tomohiro
Kuwano Hiroyasu
Dorvil Richemond
Gopstein Israel
Matsushita Electric - Industrial Co., Ltd.
LandOfFree
Method and apparatus of speech recognition and speech... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus of speech recognition and speech..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus of speech recognition and speech... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2596274