Convolutive Blind Source Separation Based on Disjointness Maximization of Subband Signals

Abstract

Abstract-The concept of disjoint component analysis (DCA) is based on the fact that different speech or audio signals are typically more disjoint than mixtures of them. This letter studies the problem of blind separation of convolutive mixtures through the subband-wise maximization of the disjointness of time-frequency representations of the signals. In our approach, we first define a frequency-dependent measure representing the closeness to disjointness of a group of subband signals. Then, this frequency-dependent measure is integrated to form an objective function that only depends on the time-domain parameters of the separation system. Lastly, an efficient natural-gradient-based learning rule is developed for the update of the separation-system coefficients.Index Terms-Convolutive blind source separation, disjointness maximization.

Secondary Surveillance Radar replies source separation via the Disjoint Component Analysis

“…The DCA criterion proposed by Annemüller has been developed later for disjoint and quasi-disjoint signals as in [12] and in several other areas such as the acoustic source localization [13], low rank decomposition [14] where DCA allowed the separation between the horizontal and vertical components of seismic images where the other algorithms could not solve this problem. The "disjointness" constraint was also used in the non-negative matrix factorization (NMF) [15] in its time-frequency form [16], and in [17] for convolutive signals.…”