EMD AND HURST BASED MODE SELECTION FOR SPEECH ENHANCEMENT
Anuja P Soman · 2015
Abstract-This paper presents a speech sweetening technique for signals corrupted by non stationary acoustic noises. The pro-exhibit approach applies the empirical mode decomposition (EMD) to the rip-roaring speech signal and obtains a set of intrinsic mode functions (IMF). the most contribution of the projected procedure is that the adoption of the Hurst exponent within the selection of IMFs to reconstruct the speech. This EMD and Hurst-based (EMDH) approach is evaluated in speech sweetening experiments considering environmental acoustic noises with totally different indices of non stationary. The results show that the EMDH improves the segmental S/N Associate in Nursing and overall quality composite live, encompassing the sensory activity analysis of speech quality (PESQ). Moreover, the short-time objective understandability (STOI) measure reinforces the superior performance of EMDH. Finally, the EMDH is additionally examined in an exceedingly Speaker Identification task in rip-roaring conditions. The projected technique ends up in the highest Speaker Identification rates when compared to the baseline speech sweetening algorithms and conjointly to a multi condition coaching procedure.