Noise reduction in multichannel neural recordings using a new array wavelet denoising algorithm

We investigate a new technique for noise reduction in multichannel neural r ecordings based on the discrete wavelet transform. Starting with the denois ing technique proposed by Donoho et al. (IEEE Trans. Inform. Theory 41 (199 5) 613-627), we suggest a new thresholding method for the multiresolution d ecomposition of the multichannel data. The potential of this technique lies in the fact that thresholds at different resolution levels of the wavelet transform are estimated spatially to account for significant correlation of the wavelet coefficients across channels. The method is applied to a simul ated multichannel data as well as real silicon microprobe recordings obtain ed in our laboratory. Preliminary results show the ability of the technique to reduce both spatially correlated and uncorrelated noise components in t he neural recordings. Results are compared to existing techniques and the o verall performance is evaluated. (C) 2001 Published by Elsevier Science B.V .