Feature selection and recognition of electroencephalogram signals: An extreme learning machine and genetic algorithm-based approach

Abstract

The effective recognition approach of the electroencephalogram (EEG) signals can significantly boost the performance and the development of the EEG-based diagnosis and treatment. A new approach which combines the Extreme Learning Machine (ELM) with the Genetic algorithm (GA) is proposed in this paper. In the proposed approach, the ELM is used both as the final classifier and the fitness function for the GA to select the optimal feature subset from the initial features extracted through time-frequency (TF) analysis. The GA is adopted as the complementary input optimization mechanism to improve the performance of the ELM. To testify the performance of the proposed approach, experiments were simulated using the real-world EEG signals of 2003 International BCI Competition dataset. The recognition results have proved the effectiveness of the proposed approach.