Phone based voice activity detection using online Bayesian adaptation with conjugate normal distributions

In this paper, we developed a highly efficient frame-level online adaptive voice activity detection (VAD) algorithm for the telephone-based CU Communicator spoken dialog system. The adaptive algorithm uses prior speaker and channel statistics as well as acoustic features of current sample frames to update model parameters. The algorithm achieved .05xRT in contrast to .7xRT of a compared VAD algorithm using 5-state HMMs. We detail the adaptive algorithm and address some real-time implementation issues. Experiments on live collected data show that there is a 23% error reduction compared with G.729B VAD.