EMD-Based Noise-Robust Method for Speech/Pause Segmentation

Abstract

The article presents a noise-robust method for speech/pause segmentation based on empirical mode decomposition. The method has been developed on the basis of a combined analysis of zero-crossing rate and short-term energy using empirical mode decomposition at the stage of preprocessing. Based on the results of preliminary processing, a set of new investigated signals, containing the most reliable information about the boundaries of the beginning and the end of informative sections of noisy speech, has been formed. The effect of the decomposition method and the influence of fragment duration of the investigated signals on the segmentation efficiency of noisy speech at different signal-to-noise ratio levels, from 20 to -5 dB with a step size of 5 dB, were assessed. The research results have shown a decrease in the values of the first and second kind errors during the segmentation of noisy speech signals using the proposed noise-robust method.