A genetic algorithm for the multidimensional filtering process of visual evoked potential signals
Nikolaos Laskaris, Anastasios Bezerianos, S. Fotopoulos, Panagiotis Papathanasopoulos · 2002
A new filtering approach for the recovering of visual evoked potential signal is introduced. The single trials are treated as observation vectors from a multidimensional distribution, in order to exploit the obvious correlation between the time instants. The technique of potential functions is utilized to approximate the vector distribution. A genetic algorithm is employed for the location of the mode of this distribution which is proposed as an estimate of the underlying EP signal. The experimental results indicate that the mode of the vector distribution deviates a lot from the corresponding ensemble average, justifying our approach. The proposed filtering method offers significant increases in SNR compared to conventional averaging, showing that the status of the visual pathway can be better portrayed.