Source separation based on a double affine projection algorithm

In a variety of contexts, observations are made of the outputs of an unknown multiple-input multiple-output linear system from witch it is of interest to identify the system to recover its input signals. This often arises, for example, in adaptive noise cancellation (ANC), and in data communications in presence of cross-coupling effect between the transmission channels. We consider the separation of two statistical independent signals using the signals captured by two sensors. The application of an adaptive filter often requires the relative placement of the two sensors at a distance that necessitates a large order filter. In this paper a symmetric feedback implementation scheme that permits a closer placement of the sensors and allows the cancellation of interference in the presence of crosstalk is used. We consider the coupling systems modeled as linear time-invariant finite impulse response (FIR) filters and we propose a new recursive-based adaptive filter solution. The optimum filter weight adaptation is based on a double affine projection algorithm (DAPA). A comparative study with other adaptive algorithms shows the superiority of the DAPA in performance.