Wavelet decomposition of myoelectric reflexes--Time feature extraction of the blink reflex and its use in clinical diagnosis

Abstract

Analysis of clinical electrophysiological signal involves noise reduction, feature extraction and classification. In evoked responses, the time latencies of electrical events provide information about the underlying physiology. Time frequency methods retain time information allowing frequency selectivity. We used the Daubechies-4 wavelet on the late blink reflex to extract time features from frequency bands where signal-to-noise ratio is the highest. This method proved to be robust in discriminating patients with multiple sclerosis from normal subjects.