EMD-Based Speech Denoising with Instantaneous Noise Modeling
This paper introduces a fully data-driven speech denoising framework based on Empirical Mode Decomposition (EMD) combined with instantaneous, frame-dependent noise modeling. The proposed method departs from classical EMD-shrinkage approaches by (i) estimating noise directly from the noisy signal without requiring voice...