Sleep Stage Scoring of Single-Channel EEG Signal based on RUSBoost Classifier

Automatic sleep stages detection in medical applications using an intelligent method is one of the most important issues in recent years which leads to reducing the workload of specialist in analyzing sleep data with visual inspection. In this paper, a single-channel EEG-based algorithm is presented for automatic recognition of sleep stages using complete ensemble empirical mode decomposition (CEEMD), hybrid genetic algorithm (GA) and neural network. First, the EEG signal is decomposed into IMFs using CEEMD and statistical and temporal features are extracted. For dimensionality reduction of feature space, a hybrid GA and multi-layer back propagation neural network is applied. Then, McNemar's test is used to validate the features. The RUSBoost algorithm performs the final classification on these optimized features and on average, the accuracies of classification for 2 up to 6 classes are 98.10%, 95.12%, 93.09%, 90.16% and 91.0% and Kappa Cohen coefficients are 0.98, 0.95, 0.93, 0.89 and 0.91, respectively. It is shown that the proposed method has better performance in classification of sleep stages compared to the previous studies.