Method, apparatus, computer system and storage medium for speech recongnition

摘要

The present invention provides a method and apparatus for speech recognition. The apparatus of the invention can include an acoustic processor. The acoustic processor can convert analog speech input signals into digital signals. Also included can be a first storage structure which can store an acoustic model which has learned voice characteristics. A second storage structure which can store a dictionary containing a first language model which has been trained regarding disfluency words and non-disfluency words, and a second language model which has been trained regarding non-disfluency words and trained to ignore disfluency words can be included. Additionally, the apparatus of the invention can include a probability calculator which can calculate a probability regarding the digital signals using the acoustic model and the dictionary to recognize words showing the highest probability of representing the input signals.