Research On Speech Recognition And Speech Command System Based On Wheel-chair Robot

Concluded from the development of robot and speech recognition technology, the research combined speech recognition technology and robot technology has been focus. It is Human Computer Interaction which is used on Welfare robot that is this kind well-knit practical research. It has not only theoretic significance but also biggish practical value. For example, handicapped can command wheelchair by his voice.Although many research work have been done on speech recognition and have made a lot of progress. Generally, It remains some following difficulties in practical application:(1) Extreme point detecting: usually, speech samplers are continuant, there no pause between words. It makes model matching difficult. Even in quiet circumstance, more than 50% errors may caused by extreme point detecting model.(2) Changeability of speech: For same word, different people have different pronunciation. Even one people will pronounce differently at two times.(3) Illegibility of speech: some different words sound similar especially in Chinese. This kind of illegibility of speech require detail resolving power during realtime speech recognize module.For these problems, this paper discuss following resolvents:(1) This paper brings forward a new process extreme point detecting arithmetic based on two kinds speech characters. It's feasibility has been proved by experiments.(2) Analyzing main limitations of DTW, this paper brings forward "DTW with coboundary and known quantity". It can make up traditional DTW's shortcomings.(3) For settling the problem caused by illegibility of speech, this paper brings forward hierarchical discriminance arithmetic during realtime speech recognize module. It focuses detail's differences and improves the system's recognition veracity and responds speed.