Time Domain Blind Separation of Nonstationary Convolutively Mixed Signals

We propose a new algorithm for solving the Blind Signal Separation (BSS) problem for convolutive mixing completely in the time domain. The closed form expressions used for first and second order optimization techniques derived in [1] for the instantaneous BSS case are extended to accommodate the more practical convolutive mixing scenario. Traditionally convolutive BSS problems are solved in the frequency domain [2--4] but this requires additional solving of the inherent frequency permutation problem. Where this is good for higher order systems, systems with a low to medium number of variables benefit from not being subject to a transform such as the DFT. We demonstrate the performance of the algorithm using two optimization methods with a convolutive synthetic mixing system and real speech data.