AN IMPROVED METHOD FOR ROBUST SPEECH ENDPOINT DETECTION

摘要：

An improved method based on adaptive band-partitioning spectral entropy for robust speech endpoint detection is provided. Adaptive band-partitioning spectral entropy which is extended from the spectral entropy is a new method for improving the robustness of speech endpoint detection. Although it has good robustness, the accuracy also degrades when the signal-to-noise ratios (SNRs) are low. Therefore, an improved solution is proposed in this paper. It utilizes adaptive filter to boost up the SNRs. The implementation procedure is given in detail. It is shown that the new algorithm can operate effectively under changeable SNRs.

关键词： Speech endpoint detection Adaptive filter Adaptive band-partitioning spectral entropy Robustness

作者: HAI-NAN LONG CUI-GAI ZHANG

作者单位: College of Electronic and Information Engineering, Hebei University, Baoding 071002, China

会议类型: 国际会议

会议名称: 2009 International Conference on Machine Learning and Cybernetics(2009机器学习与控制论国际会议)

会议地点: 保定

会议语种:英文

页码: 2067-2071

在线出版日期: 2009-07-12（万方平台首次上网日期，不代表论文的发表时间）