Speech signal noise reduction by EMD

In this paper, a speech signal noise reduction based on a multi-resolution approach referred to as Empirical Mode Decomposition (EMD) [1] is introduced. The proposed speech denoising method is a fully data-driven approach. Noisy signal is decomposed adaptively into oscillatory components called Intrinsic Mode Functions (IMFs), using a temporal decomposition called sifting process. The basic principle of the method is to reconstruct the signal with IMFs previously thresholded using a shrinkage function. The denoising method is applied to speech with different noise levels and the results are compared to wavelet shrinkage. The study is limited to signals corrupted by additive white Gaussian noise.