Genetic algorithm based ensemble system using MLR and MsetCCA methods for SSVEP frequency recognition. (January 2023)

• Record Type: Journal Article
• Authors: Ziafati, Amir; Maleki, Ali
• Is Part Of: Medical engineering & physics. Volume 111(2023)
• Publication Date: 2023-01
• Subjects: Brain-computer interfaces -- SSVEP signals -- Multiple linear regression -- Multivariate canonical correlation analysis -- Genetic algorithm (GA) -- Ensemble method; Biomedical engineering -- Periodicals; Biomedical Engineering -- Periodicals; Physics -- Periodicals; Génie biomédical -- Périodiques; Biomedical engineering; Electronic journals; Periodicals; 610.28
• DOI: 10.1016/j.medengphy.2022.103945
• Languages: English
• ISSNs: 1350-4533

Abstract

Highlights: MLR and MsetCCA are the most powerful methods to detect SSVEP stimulation frequency. GA-based ensemble system yields better results than MLR and MsetCCA methods. The results of the proposed ensemble system show 100% accuracy in 2-second signals. The GA-Ensemble method produces the best results without any expert intervention. Abstract: BCI systems provide a direct communication channel between the human and the machine using brain signals. Among the various methods of steady-state visual evoked potential (SSVEP) stimulation frequency detection, multiple linear regression (MLR), and multiset canonical correlation analysis (MsetCCA) methods have achieved high accurate results in recent studies. The purpose of this study is to utilize both approaches and benefit from them using a genetic algorithm (GA). This algorithm leads to high-performance optimization due to its large number of regulatory parameters. Signal analysis was performed for the windows with 0.5 to 4 s duration length and with 0.5-second incremental steps. In this paper, we were able to achieve 100% accuracy of recognition for 2-second time-windows using the genetic algorithm to optimally ensemble SSVEP stimulation frequency detection methods. The accuracy of the proposed system indicates a significant improvement in detection compared to either MLR or MsetCCA alone and indicates that the ensemble system is correctly optimized using the genetic algorithm. Genetic algorithm is one of the most widely used algorithms because of its high regulatory parameters leading to its high flexibility. The improvement in detection of the proposed system is due to the use of the strengths of both two methods, and the optimal choice of the system response to visual stimuli.