Ensemble classifier based on optimized extreme learning machine for motor imagery classification

Designing an effective classifier with high classification accuracy and strong generalization capability is essential for brain-computer interface (BCI) research. In this study, an extreme learning machine (ELM) based method is proposed to improve the classification accuracy of motor imagery electroencephalogram (EEG).

The proposed method constructs an ensemble classifier based on optimized ELMs. Particle swarm optimization is used to simultaneously optimize the input weights and hidden biases of ELM to avoid the randomness and instability of classification result when ELM uses randomly generated parameters, and majority voting strategy is used to fuse the classification results of multiple base classifiers to avoid the negative impact of ELM with local optimal parameters on classification result. The proposed method was compared with four competing methods in experiments based on two public EEG datasets and some existing methods reported in the literature using the same datasets as well.

The results indicate that the proposed method achieved significant higher classification accuracies than those of the competing methods on both two-class and four-class motor imagery data. Moreover, compared to the existing methods, it still obtained superior average accuracies of two-class classification and performed better for the subjects with relatively poor accuracies on both two-class and four-class classifications.

The significant accuracy improvement demonstrates the superiority of the proposed method. It can be a promising candidate for accurate classification of motor imagery EEG in BCI systems.