Statistical Properties of Stft Ratios for Two Channel Systems and Applications to Blind Source Separation

The ratio of the short time Fourier transform (STFT) coe cients of signals received at two sensors can factor out the role of the power spectrum of emitting sources, under an assumption called disjoint orthogonality. Thus, it can reveal parameters speci c to the mixing scenario and serve as a basis for channel estimation techniques. In this paper we analyze and extend a source separation method based on the use of STFT ratios of two sensor inputs, called DUET. We generalize the problem formulation in DUET, and prove that considerably weaker assumptions about the classes of input signals are su cient to apply the derived techniques. The analysis centers on the notion of ratio-estimator and a novel stochastic model that enabled us to derive the maximum likelihood ratio-estimator. Derived techniques can be e ectively applied to source separation, source localization, signal enhancement, and noise reduction when using a twin microphone system, both in echoic environments and degenerate situations.