In this paper we propose a new algorithm of phoneme based speaker-independent English command recognition and developa speaker-independent English command recognition system. It accelerates the whole system development by using HTK (hide Markovtoolkits) and Visual C + + based on the characteristics of speaker-independent speech recognition. In recognition phase we combinethe confidence measures with incomplete matching, which considerably improve the quality of recognition. The recognition accuracyis increased by 4.8% over complete matching without back-end processing when the sige of vocabulary is more than 10.