Voice Activity Detection Based on Chi-square Distribution

摘要：

In this paper, we develop a voice activity detection (VAD) algorithm based on chi-square distribution. The noisy speech is decomposed into bark-domain band signals and the distributions of sub-band signals are modeled by chi-square distributions. The algorithm is based on the determination of the speech/noise divergence by means of log-likelihood ratio of the sub-band signals probability density. The proposed VAD is tested on speech signals with various noises under different SNRs. According to experimental results, it is found that the proposed VAD outperforms the standard algorithm G.729 VAD.

关键词： chi-square Voice Activity Detction (VAD) Bark Domain

作者: Wu Qi-qian Zou Xia Zhang Xiong-wei

作者单位: Institute of Communication Engineering PLA University of Science and Technology Nanjing, China Institute of Command Automation PLA University of Science and Technology Nanjing, China

会议类型: 国际会议

会议名称: 2010 International Conference on Circuit and Signal Processing(2010年电路与信号处理国际会议 ICCSP 2010)

会议地点: 上海

会议语种:英文

页码: 84-87

在线出版日期: 2010-12-25（万方平台首次上网日期，不代表论文的发表时间）