Method of recognizing coherently spoken words

摘要: During the recognition, speech values which are derived from sample of signals compared with reference values, words a given vocabulary each time being by sequence values. The then determined phonemes according to fixed pronouncing lexicon and for in learning phase, phoneme within word consisting number equal phase. In order approach transitions between phonemes, may also consist three sections constant By per phoneme, duration can be simulated more accurately. Different possibilities indicated determine distance value during recognition.