Mandarin Connected Digit Speech Recognition Based On Hidden Markov Models

Connected digit speech recognition is a crucial branch of continuous speech recognition .It has attracted the attention of many researchers for a long time, because it has wide applications, such as internet, communication, military affairs, human-computer dialogue system, etc.Mandarin connected digit speech recognition is more difficult than English digit recognition because of its own pronunciation characteristic.The ultimate goal of this thesis is to improve the connected digit speech recognition performance based on CHMM. Towards this goal, we investigate a variety of ways of utilizing prosodic information. Experiments show that the proposed algorithm can improve the connected digit speech recognition performance.Each unit of a speech recognition system is studied in this thesis, which includes the preprocessing, the extracting of acoustic features, the training of HMM, the separating and recognition of connected speech with prosodic information, etc. The all algorithms are realized on the software VC++. On the other hand, we examine isolated digit speech recognition before investigate connected speech recognition, and we also examine speech recognition based on DHMM before investigate recognition based on CHMM.