Noise-robust speech signals processing for the voice control system based on the Complementary Ensemble Empirical Mode Decomposition

Abstract

Noise-robust speech signals processing is one of the main problems of practical realization of voice control systems (VCS). The offered algorithm of noise-robust processing represents speech signals filtering (voice commands) with the use of the Complementary Ensemble Empirical Mode Decomposition (CEEMD) and the Independent Component Analysis (ICA) methods. A noisy speech signal is adaptively decomposed into frequency components -intrinsic mode functions (IMF) by means of the CEEMD method. The application of the CEEMD method for signals decomposition allows excluding mixing of IMF arising when processing signals containing short-term and disparate in scale areas. From the received set of IMF the mode is defined containing the main noise by means of an assessment of weight energy and noise IMF coefficients. Further the initial noisy speech signal and IMF with the main noise are exposed to processing by means of the ICA method. As a result the filtered speech signal is allocated. The application of the offered filtering algorithm contributes to the increase of VCS noise resistance and accuracy of voice commands recognition. The results of the offered algorithm researches show the effective noise suppression, including small values of signal-to-noise ratio (SNR).