Use of Microphone Array and Model Adaptation for Hands-Free Speech Acquisition and Recognition

摘要: This paper presents a combined microphone array and model adaptation algorithm for hands-free speech recognition. Our purpose is to remove the inconvenience of using head-mounted/hand-holding in conventional recognizer. To improve quality with car noise interference, linear applied acted as robust acquisition system. A time-domain coherence measure (TDCM) reliably estimate time delay signals collected by different microphones. The estimated adopted delay-and-sum beamformer enhancement. Further, we adapt hidden Markov models get close acoustic conditions enhanced test In recognition experiments connected Chinese digits, found that TDCM can effectively delay. increase sampling rate helpful determine Incorporating scheme significantly reduces errors moderate computation overhead.