A Robust PCA-SURE Thresholding Deep Neural Network Approach for Mental Task Brain Computer Interface

Electroencephalographic (EEG) characteristics, i.e., non-linear structure and nonstationarity, make mental state recognition not a trivial task to various classification models. In this paper, a combined principal component analysis (PCA) – Deep neural network  approach is proposed as a robust and effective solution to classifying EEG signals recorded with low-cost and portable recording systems into different mental states towards implementation of a Brain computer interface  (BCI) capable of controlling electronic devices. Stein’s unbiased risk estimate (SURE) thresholding – PCA is utilized to obtain spectral features that are most essential for deep neural network classifier to perform at best. The contributions of this paper are three–fold. First, we propose a novel, robust and efficient method that integrates SURE thresholding, PCA and Deep Neural Network (DNN), along with other signal processing and machine learning techniques for mental state recognition. Second, a complete mental-task-based BCI using an appropriate experimental design without any assistant equipment is presented. Third, SURE risk thresholding is utilized and proven to be an effective method to automatically determine the appropriate number of principal components of EEG features returned by performing PCA. Experimental results show that our method outperforms others in an EEG dataset of four subjects with highest classification accuracies on dual and triple mental state task experiments of 96.83% and 76.90%, respectively.