In this paper, Chinese language speaker-independent continuous speech recognition technology got deep, systematic study, and on this basis, I realize an independent continuous speech recognition system. This article mainly includes the following sections:(1) it describes the endpoint detection and its application, a detailed study of the mainstream feature extraction methods are also involved, and I also study how to get more robust feature, which has higher tolerance to noise in the environment. (2) I carefully study the acoustic modeling, and according to Chinese languageās particularities, the acoustic model is got some special treatment. (3) I also have a carefully study of the language model, and for this purpose of speech recognition, I put some of my own idea into the language model. (4) I fulfill the continuous recognition system, and in the meanwhile I have got a series of valuable conclusion, and finally I make some prospect of my future study in speech recognition. |