Classification of Cognitive Activities EEGs via Singular Spectrum Analysis and Joint Approximate Diagonalization of Eigen-Matrices Algorithm with Ensemble Classifier

Cognition involves the recognition and perception of external objects to solve creative problems and adapt to the environment, forming the foundation of human intelligence. However, neurodegenerative diseases such as Alzheimer's disease (AD) and Parkinson's disease (PD) lead to cognitive impairment. Transcutaneous electrical nerve stimulation (TENS) is a recognized therapy for enhancing cognitive performance. This paper proposes using singular spectrum analysis (SSA) and the joint approximate diagonalization of eigen-matrices (JADE) algorithm with an ensemble classifier to classify electroencephalograms (EEGs) before and after TENS application during various cognitive activities. Initially, subjects wear a headband to acquire single-channel EEGs before and after TENS is applied to their radial nerves. Empirical mode decomposition (EMD) is then used to suppress underlying EEG trends. Subsequently, SSA and the JADE algorithm perform blind source separation (BSS) of EEGs from electromyograms (EMGs) and electrocardiograms (ECGs). JADE is a fourth-order based independent component analysis (ICA) algorithm. The tunable Q-factor wavelet transform (TQWT) decomposes EEGs into subbands from which features are extracted. Finally, an ensemble classifier, combining random forest and least squares support vector machine (LS-SVM), classifies the data. The computer numerical simulation results demonstrate that our proposed method achieves higher classification accuracy compared to existing methods.