Blind source separation of nerve cuff recordings

Nerve cuff Electrodes record an aggregate signal from a nerve comprised of several fascicles. Recovering individual fascicular signals could be important for the control of prosthetic devices. Blind Source Separation methods have been designed for recovering individual source signals from multi-channel recordings of a mixture of multiple sources. Here, we present a simulation study of the feasibility of applying Blind Source Separation (BSS) to recover fascicular signals from multiple contact nerve cuff electrodes. Spontaneous neuroelectric activities and their recordings are simulated. Hyvarinen's FastICA algorithm, a popular method for BSS, combined with a correlation analysis is used for separation. The ability of the method to separate the fascicular signals was estimated by measuring the correlation between the input and output signals. In 25 random trials with different signals the mean value of the correlation coefficient was 0.86 (/spl plusmn/ 0.15 S.D.). Although all signals were recovered every time, the algorithm was not able to determine the origin of the signals. The BSS procedure permutes separated signals randomly.