Appeal No. 1997-4121
Application No. 08/427,272
employed in such speech recognition systems is typically prepared by using the word feature vectors
obtained by using specific word boundaries and a specific noise level. Therefore, when the conventional
dictionary is used with the word spotting method, matching with the word feature vector obtained from an
unfixed word boundary for a speech mixed with noise having a low signal/noise ratio, as in a practical
environment, causes some recognition errors. The instant invention is directed to obtaining a high
recognition rate even in noisy environments. It also includes an effective learning system for word spotting
methods of speech recognition.
Representative independent claim 1 is reproduced as follows:
1. An apparatus for time series signal recognition, comprising:
means for inputting signal patterns for time series signals to be recognized;
means for recognizing the time series signals according to a word spotting scheme using
continuous pattern matching, including;
means for extracting a plurality of candidate feature vectors for characterizing an
individual time series signal from the signal patterns;
recognition dictionary means for storing reference patterns with which the individual
time series signals are matched;
means for calculating similarity values for each of the extracted candidate feature
vectors and the reference patterns;
means for determining a recognition result by selecting one of said stored reference
patterns that matches with one of the candidate feature vectors by the continuous pattern matching
for which the similarity value calculated by the calculating means is greater than a prescribed
threshold value; and
means for learning new reference patterns to be stored in the recognition dictionary
means, including:
means for mixing speech patterns with noise database patterns representing
background noises, to form signal patterns for learning, and supplying the signal patterns for learning to the
recognizing means;
means for extracting feature vectors for learning from the recognition results and the
2
Page: Previous 1 2 3 4 5 6 7 8 9 Next
Last modified: November 3, 2007