Motor-imagery BCI task classification using Riemannian geometry and averaging with mean absolute deviation

Abstract

Brain Computer interface (BCI) is thought as a better way to link within brain and computer alternative machine. Many types of physiological signal will work BCI framework. Motor imagery (MI) has incontestable to be a excellent way to work a BCI system. Recent research concerning MI based mostly BCI framework, lower performance accuracy and intense of time have common issues. Main focuses of this paper is select the appropriate central point of tangent space in Tangent Space Linear Discriminant analysis-based Motor-Imagery Brain-Computer interfacing. Method name tangent space mapping LDA (TSMLDA) analysis takes its moves from the observations that normally, the EEG signal embodies outliers, so the centrality as a geometric mean of tangent space might not be the simplest alternative. We tend to propose the employment of strong estimators of variance matrices average. Specifically, Median Absolute Deviation(MAD) going to be planned and mentioned. Associate in Nursing experimental analysis can show the advance of Tangent house Linear Discriminant Analysis corresponding to the planned strong estimators. Experimental results show that our proposed method performs 3% better than the recently developed algorithms. © 2019 IEEE.