Speech Endpoint Detection Based on Improved Adaptive Band-partitioning Spectral Entropy
Ning Jiang · Jisuanji fangzhen · 2008
Accurate Speech endpoint detection in adverse environments is very important for robust speech recognition.Adaptive band-partitioning spectral entropy is a new method for improving the robustness of speech endpoint detection. The idea of the method is to divide a frame into some sub-bands which the number of it could be selected adaptively,and calculate spectral entropy of them.Adaptive band-partitioning spectral entropy method is extended from the spectral entropy.Although it has good robustness,the accuracy degrades rapidly when the SNRs are low. Therefore,an improved adaptive band-partitioning spectral entropy was proposed for speech endpoint detection,which utilized the weighted power spectral subtraction to boost up the SNR as well as keep the robustness. The implementation procedure was given in detail. The speech recognition experiment results indicate that the recognition accuracy has improved well in adverse environments.