Bio-inspired Red Fox-Sine cosine optimization for the feature selection of SSVEP-based EEG signals for BCI applications

Abstract

BackgroundAdvancements in Brain-Computer Interface (BCI) have led to the development of various neuro-dysfunctional human assistive tools. Despite having various human-assistive tools, communication is one of the prime barriers for paralytic people. Preliminary investigations to resolve the communication problem of the paralytic people have revealed that the Steady State Visually Evoked Potential (SSVEP) signal classification has the potential to decode the needs of the paralytic people. In that, optimal feature selection of SSVEP-based EEG signals has a dominant role to make the signal classification more effective in terms of accuracy and computation time. The optimal feature selection is one of the critical and time-consuming tasks hence it has been formulated as an optimization problem. New MethodsTo select the optimal features subset a hybrid Red Fox and Sine-Cosine Optimization algorithm (RFO_SCA) is proposed in this paper. The error minimization function of the classifier is used as a fitness function for RFO_SCA. For local optimal subset selection, Sine-Cosine Algorithm (SCA) is employed to search for the best local optimal solution. This strategy improves the local search efficiency of Red Fox Optimization (RFO). The selected optimal features are further classified using k-NN, decision tree, and random forest classifier. ResultsTo validate the efficiency of RFO_SCA in feature optimization, the experiment has been carried out in two datasets such as (i) the acquired dataset and (ii) EEG SSVEP dataset III from the MAMEM database. The experimental results show that RFO_SCA outperforms other optimization algorithms with an accuracy of 98.74% and 92.14% for the acquired and standard datasets respectively. Also, the feature selection using RFO_SCA has reduced the feature size by 50% and shows a high convergence rate when compared to other optimization algorithms.