Voice activity detection based on ensemble empirical mode decomposition and teager kurtosis

This paper proposes an improved voice activity detection (VAD) methodology based on ensemble empirical mode decomposition (EEMD) algorithm and the teager kurtosis to avoid the defect of empirical mode decomposition (EMD) in mode mixing. The teager energy operator is used to track the modulation energy of each intrinsic mode function (IMF), decomposed by ensemble empirical mode decomposition. The root power function and order statistics filter are used on the teager kurtosis for feature extraction. Voice activity detection can be implemented over the suitable threshold which can be automatically estimated by tracking the minimum of the extracted feature values. Experiments show that the proposed VAD can achieve comparable results at high signal-to-noise ratio (SNR). For low SNR conditions, it is able to maintain lower error detection ratio and higher detection ratio, compared with those of the original algorithm.