Speech recognition technology has achieved significant progress with many researcher's enormous efforts in the past tens of years. In current years, electronic products have developed rapidly, and speech recognition technology has great application potential as one of convenient interactive means of human and machine. Speaker independent isolated word recognition for embedded systems has becoming one of hotspots of current research. In the paper we focus our research on high performance and low consume Speaker independent isolated word recognition algorithm. We also discuss several questions in practical recognition system. There are several points in my work: 1. Develop a new Speaker independent isolated word recognition algorithm based on basic acoustic unit modeling and whole word matching. Study the performance and consume of the algorithm in different feature parameters and model parameters. We also research the application of duration information in speech recognition. 2. Research speaker adaptation technology in isolated word recognition, discuss two kind of common adaptation algorithms and develop a new SATD algorithm for our isolated word recognizer for embedded system. 3. Because the existence of OOV words makes bad effect on the performance of speech recognizer, we study OOV rejection approaches. We compare the difference of OOV rejection approaches of speaker dependent and speaker independent system. We finally select some effective approaches for our recognizer. 4. Realize our Speaker independent isolated word recognition algorithm on DSP platform. Introduce the hardware structure and software module design of the DSP system. We also evaluate the whole performance of the recognizer using new algorithm.
修改评论