Information-theoretic approach to multichannel signal extraction by multiple interference cancellation in electrochemical array data

An information-theoretic approach oriented to develop on-line algorithms for electrochemical array data processing is presented. We deal with the multichannel processing of the signals acquired by an array of silicon-based chemical sensors. The objective is to extract multiple desired signals by the cancellation of multiple interferences in high noisy environments. The nonlinear mixture separation is achieved in the presence of interference and strong cross nonlinearities, by minimizing the output mutual information in multi-input, multi-output learning machines, considering mutually independent sources. Numerical results demonstrate the viability of the proposed approach in the context of array signal extraction.