A Novel Approach to Denoise Visual Evoked Potentials Using Higher Order Statistics

Evoked Potentials (EPs) are time varying signals buried in large background noise. In order to detect and extract Visual Evoked Potentials (VEP) in real time, we use the fourth order cumulant of the observed noisy signal in an adaptive filter. The noisy signal is passed through a Finite Impulse Response (FIR) filter whose impulse response is matched with the shape of the noise-free signal. The impulse response of the said filter is estimated using the fourth order cumulant of the input signal. We also use a method to recursively utilize fourth order cumulants of the input signal for updating the coefficients of the adaptive FIR filter. This enables us to extract the VEP signal in real time. We show that the fourth order cumulant based method provides better results in comparison to the conventional third order cumulant based method and both yield better results as compared to the widely used autocorrelation function.

Keywords:

Visual Evoked Potentials , Higher-Order-Statistics , Adaptive Methods , Colored Gaussian Noise , Fourth-Order Cumulants