Voice Activity Detection Based on EMD and Power Spectrum Entropy
Wang Hu · Audio Engineering · 2013
In order to improve the accuracy of voice activity detection in low SNR environments,a new method based on the empirical mode decomposition and power spectrum entropy is proposed to identify speech-segment endpoints. Noisy speech signals are decomposed into a set of intrinsic mode functions,take power spectrum entropy as the feature of voice activity detection,and compute power spectrum entropy of IMFS to achieve voice activity detection. The method of empirical mode decomposition can effectively eliminate the disturbance of additive white Gaussian noises. The simulation results show that combine the methods of power spectrum entropy with empirical mode decomposition can achieve voice activity detection effectively in low signal to noise ratio environments.