Multi-sensor asynchronous information fusion classification strategy based on coupled HMM
Jizhong Liu · Jisuanji yingyong yanjiu · 2009
To solve the problem of multi-sensor asynchronous information fusion,this paper proposed a new fusion strategy based on coupled HMM.This approach not only realized time correlation,but also solved the information asynchronism among multi channels.Second,simplified the model structure by confining state numbers and limitting the asynchronism.Furthermore,to realize algorithm of coupled HMM,represented CHMM by a equivalent 2-stream HMM,so traditional HMM algorithm could be directly used on CHMM.The experiments performed on the fusion of speech and lipreading have shown that this strategy achieves better results than early synchronous fusion strategy.