LNC UMR 6155 Universit de Provence Marseille ESCAN

LNC - UMR 6155 Université de Provence - Marseille ESCAN 2012 1 st conference of European Society of cognitive and Affective Neuroscience Estimation of the Individual Evoked Potential by Wavelet filtering and Bootstrap method Moncef Benkherrat, Franck Vidal, Thierry Hasbroucq, Boris Burle Abstract – A new method to improve the signal-to-noise ratio of single evoked related potentials (ERPs) measurements is presented, in which, contrary to previous methods, no a priori assumptions on the signal are necessary. This method is based on the wavelets decomposition of the individual signals. A statistical thresholding is applied on the coefficients of the decomposition: we estimate whether the mean value of the coefficients across trials and for each time point is significantly different from a random estimate. The performance of the method is evaluated against similar ones with simulations and the method is applied to real data. Event Related Potentials Wa. SDe Method 1 - Wavelet decomposition Stimulus Bra in Multiresolution Wavelet Transform 2 - Thresholding of wavelet coefficients : 2 -1 - estimation of the empirical distribution, 2 -2 - determine the inferior threshold inf and the superior threshold sup corresponding to a confidence set at 0. 05, 2 -3 - apply the thresholding rule : Decomposition Comparison of the three methods by the calculation of the mean square error for various SNR Real Data Processing the wavelet coefficients whose significance is below to 0. 05 confidence interval are kept while the other coefficients are set to zero. 3 - The inverse wavelet transform is applied to obtain the denoised signal. §Corresponds to a downsampling by a Simulation factor of 2. H and G are respectively the decomposition low and high pass filters. ERPs before denoising Reconstruction The wavelet reconstruction process consists of upsampling and filtering. RH and RG are respectively the reconstruction low and high pass filters Comparison of three methods : Wa. SDe method, Guiroga method and Wang method for SNR= -10 d. B Conclusion : We proposed a method to improve the SNR of an individual signal which can be applied without any a priori assumptions about the signal and the distribution of the wavelet coefficients. Comparison with recent methods by simulation and results obtained on real data has shown the efficiency of the proposed algorithm. ERPs after denoising with Wa. SDe method