Detection of sleep spindles by discrete wavelet transform

Reliable detection of sleep spindles from multichannel electroencephalogram (EEG) data has become an important issue recently. However, since the EEG data are usually accompanied with artifacts, pre-processing of these data has to be performed in order to increase detection accuracy. In this paper, the authors present the Discrete Wavelet Transform as an alternative method to capture the spindle activity. Wavelet Transform's advantage is its inherent way of decomposing the signal onto orthonormal dyadic frequency bands, thus excluding the need for notch or band pass filtering of the signal to eliminate electrical interferences. The authors display their results by a topographic map of the spindle activity.