Being the computer system in order to choose the system for skill training which was adjusted dynamically

摘要

To familiarize SUMMARY train speech recognition system (57) Method and system for dynamically selecting words. The model each phoneme using a hidden Markov model speech recognition system is represented as a sequence of phonemes of each word. According to the probability that a corresponding codeword is spoken as part of the phoneme, familiarization training system ranks each phoneme for each frame. Familiarization training system collects the utterance spoken word corresponding is known. Then, familiarization training system aligns the codeword of each utterance for phoneme code word has been recognized as a part. Using the codewords aligned for frame aligned, familiarization training system calculates the average rank of each phoneme. Finally, familiarization training system to select the familiarization training for words that contain the phoneme low rank.