摘要:
PROBLEM TO BE SOLVED: To improve identification accuracy.SOLUTION: A pattern identification device includes a reception part, a determination part, an execution part. a calculation part and a decision part. The reception part receives an input pattern and attribute information of the input pattern. The determination part determines a subclass which the input pattern belongs to on the basis of at least the attribute information. The execution part identifies whether the input pattern belongs to a class by using a weak discriminator allocated to the determined subclass and outputs an identification result and a degree of reliability of the weak discriminator. The calculation part calculates an integrated value obtained by integrating an evaluation value based on the identification result and the degree of reliability. The decision part decides whether termination conditions of identification processing by the determination part, the execution part and the calculation part are satisfied and repeats the identification processing when the termination conditions are not satisfied while it ends the identification processing and outputs an integrated value at the time of ending when the termination conditions are satisfied.
摘要:
A distance calculation unit (16) obtains the acoustic distance between the feature amount of input speech and each phonetic model. A word search unit (17) performs a word search based on the acoustic distance and a language model including the phoneme and prosodic label of a word, and outputs a word hypothesis and a first score representing the likelihood of the word hypothesis. The word search unit (17) also outputs a vowel interval and its tone label in the input speech, when assuming that the recognition result of the input speech is the word hypothesis. A tone recognition unit (21) outputs a second score representing the likelihood of the tone label output from the word search unit (17) based on a feature amount corresponding to the vowel interval output from the word search unit (17). A rescore unit (22) corrects the first score of the word hypothesis output from the word search unit (17) using the second score output from the tone recognition unit (21). This allows to raise the speech recognition accuracy for tone speech.
摘要:
PROBLEM TO BE SOLVED: To recognize an infinite number of words in principle.SOLUTION: The present invention handles a system for speech recognition, for example, for recognizing words in a continuous speech. Disclosed is a speech recognition system capable of recognizing a great number of words, or an infinite number of words in principle. The speech recognition system includes a word recognition device for deriving the best path in a word graph, and words are allocated to the speech based upon a minimum path thereof. A phoneme language model is applied to respective words of the word graph to obtain a word score. Further, the present invention relates to a device and method which identify words from a speech block, and a computer-readable code for implementing the method.