SPEECH RECOGNITION SYSTEM ABSTRACT Apparatus and method for evaluating the likelihood of a word in a vocabulary of words wherein a total score is evaluated for each word, each total score being the re-sult of combining at least two word scores generated by differing algorithms. In one embodiment, a detailed acoustic match word score is combined with an approxi-mate acoustic match word score to provide a total word score for a subject word. In another embodiment, a polling word score is combined with an acoustic match word score to provide a total word score for a subject word. The acoustic models employed in the acoustic matching may correspond, alternatively, to phonetic el-ements or to fenemes. Fenemes represent labels generated by an acoustic processor in response to a spoken input. Apparatus and method for determining word scores ac-cording to approximate acoustic matching and for deter-mining word scores according to a polling methodology are disclosed.

Speech recognition system

Speech recognition is improved by splitting each feneme string at a consistent point into a left portion and a right portion. The present invention addresses the problem of constructing fenemic baseforms which take into account variations in pronunciation of words from one utterance thereof to another. Specifically, the invention relates to a method of constructing a fenemic baseform for a word in a vocabulary of word segments including the steps of: (a) transforming multiple utterances of the word into respective strings of fenemes; (b) defining a set of fenemic Markov model phone machines; (c) determining the best single phone machine P1 for producing the multiple feneme strings; (d) determining the best two phone baseform of the form P1 P2 or P2 P1 for producing the multiple feneme strings; (e) aligning the best two phone baseform against each feneme string; (f) splitting each feneme string into a left portion and a right portion with the left portion corresponding to the first phone machine of the two phone baseform and the right portion corresponding to the second phone machine of the two phone baseform; (g) identifying each left portion as a left substring and each right portion as a right substring; (h) processing the set of left substrings and the set of right substrings in the same manner as the set of feneme strings corresponding to the multiple utterances including the further step of inhibiting further splitting of a substring when the single phone baseform thereof has a higher probability of producing the substring than does the best two phone baseform; and (k) concatenating the unsplit single phones in an order corresponding to the order of the feneme substrings to which they correspond.

Constructing Markov models of words from multiple utterances

The learning method of reference pattern vectors for speech recognition in accordance with the present invention, a plurality of speech feature vectors are generated from the time series of speech feature parameter for the input speech pattern, by taking into account knowledge concerning the variation tendencies of the speech patterns, and the learning (preparation) of reference pattern vectors for speech recognition is carried out by the use of these speech feature vectors thus generated. Therefore, it becomes possible to prepare highly reliable reference pattern vectors in an easy manner from a small number of speech patterns, which makes it possible to achieve an improvement in the speech recognition factor. In particular, it becomes possible to plan an easy improvement of the reference pattern vectors by an effective use of a relatively small number of input speech patterns.

Learning system of dictionary for speech recognition

Speech is input (11), amplified (13), digitized (15) and processed by a digital signal processor (17). A subset of stored patterns is selected by a computer (23) based on comparisons of the processed speech with the stored patterns.

Speech regognition system utilizing vocabulary model preselection.

A speech recognition system combines content analysis of a signal with directional analysis of the signal to calculate probable contents of the signal and the probability that the contents are as calculated. Content analysis can be combined with independent source direction analysis, consisting of the probable direction of the source and the probability that the source is as determined, to refine the content analysis. The direction analysis can affect either the calculation of the probable contents of the signal or the calculation of the probability that the contents are as calculated. The improved information obtained by combining the content analysis with directional analysis can be used to improve subsequent content analyses or directional analyses.

Content analyzer mixing inverse-direction-probability-weighted noise to input signal

Speech recognition executes the final matching operation for each characteristic pattern according to the preliminarily selected order prior to the final matching, where any further calculation in reference to such a characteristic pattern is not executed after a specific condition has been satisfied during the matching calculation, and as result, calculation requirements can be drastically minimized so that the speech recognition can eventually be implemented at higher speed and yet under a sharply reduced cost.

Yoshiki Hishioka

Papers

Speech recognition with preliminary matching