摘要
提出了一种噪声环境下的语音活动度(Voice Activity Detection)的稳健检测算法,算法采用了先降噪后检测的策略.为了使检测算法能够适应嘈杂的噪声环境,本文采用了两个互补性的策略.首先,采用噪声特征空间投影的方法,以较小的语音畸变为代价,去掉语音信号中的有色分量,然后利用Teager Energy Operator(TEO)来增强语音信号与噪声之间的能量差别,最终,根据子带TEO的平均信噪比来区分语音与非语音信号.我们采用了TIM IT数据库与几种常见的噪声来评价该算法,实验表明,该算法优于最新的语音活动度检测算法.
A robust voice activity detection algorithm in the presence of noise fields is proposed in this paper,which reduces noise before voice detection. In order to adapt to noise fields, the algorithm is based on two complementary strategies. Firstly, the method of noise feature space protection is adopted to reduce colored component at the expenses of small voice distortion. Secondly, Teager Energy Operator （TEO） is employed to enhance energy differences between voice and noise. The signals are then separated by average noise ratio of TEO subband signals. The evaluation of the algorithm by TIMIT database and several common noises proves that it is superior to other voice activity detection algorithms.
出处
《昆明理工大学学报（理工版）》
北大核心
2010年第3期77-82,共6页
Journal of Kunming University of Science and Technology(Natural Science Edition)
关键词
语音识别
噪声环境
去噪
TEO
speech recognition
noise environment
noise reduction
TEO
