摘要
In this paper we propose a new algorithm of phoneme based speaker independent English command recognition and develop a speaker independent English command recognition system. It accelerates the whole system development by using HTK (hide Markov toolkits) and Visual C++ based on the character'istics of speaker independent speech recognition. In recognition phase we combine the confidence measures with incomplete matching, which considerably improve the quality of recognition. The recognition accuracy is increased by 4.8% over complete matching without back end processing when the sige of vocabulary is more than 10.
In this paper we propose a new algorithm of phoneme based speaker independent English command recognition and develop a speaker independent English command recognition system. It accelerates the whole system development by using HTK (hide Markov toolkits) and Visual C++ based on the character'istics of speaker independent speech recognition. In recognition phase we combine the confidence measures with incomplete matching, which considerably improve the quality of recognition. The recognition accuracy is increased by 4.8% over complete matching without back end processing when the sige of vocabulary is more than 10.
基金
ProjectsupportedbyNationalNaturalScienceFoundationofChina(GrantNo .60072031)andTheHongKongPolytechnicUniversity (GrantNo . T499)
