A Novel Algorithm for Multichannel Deconvolutive based on αβ-Divergence

We introduce a novel Algorithm for underdetermined convolutive mixture of source signals. Where the convolution is routinely approximated in the short-time Fourier transform (STFT) domain as linear instantaneous mixing in each frequency band. Each source STFT is given a model inspired from nonnegative matrix factorization (NMF) with the -divergence, this divergence is a family of cost functions parameterized by a two tuning parameters ( and ), and smoothly connect the fundamental Alpha-, Betaand Gamma-divergences. The proposed family of -multiplicative NMF algorithms is shown to improve robustness separation with respect to noise and outliers. Our decomposition algorithm is applied to stereo audio source separation in various settings, covering blind and supervised separation, music and speech sources, synthetic instantaneous and convolutive mixtures.