Variable sampling of large-array EEG and MEG

Recent development in computer technology has allowed collection of large-array electroencephalograms (EEG) and magnetoencephalograms (MEG) consisting of hundreds of channels. As the amount of collected data increases, there is a clear demand to reduce the data size for efficient database management and network transmission. We investigated this problem and developed a simple method by eliminating data samples at smooth regions of the signal according to the signal's local frequency content. This method has two major advantages over the existing methods, (1) it reduces data size significantly with a controllable distortion, and (2) the data can be displayed directly without decompression.