Encoding polysomnographic signals into spike firing rate for sleep staging

Abstract

In this work, an encoding of polysomnographic signals into a spike firing rate, based on the BSA algorithm, is used as a discriminant feature for sleep stage classification. This proposal obtains a better sleep staging compared with the mean power signals frequency. Furthermore, a comparison of classification results obtained by different algorithms - such as Support Vector Machines, Multilayer Perceptron, Radial Basis Function Network, Naïve Bayes, K-Nearest Neighbors and the decision tree algorithm C4.5 - is reported, demonstrating that Multilayer Perceptron has the best performance.