Enhanced Nanoelectronic Detection and Classification of Motor Imagery Electroencephalogram Signal Using a Hybrid Framework

Abstract

Motor imagery-based electroencephalogram (MI-EEG) signal classification plays a vital role in the development of brain-computer interfaces (BCIs), particularly in providing assistance to individuals with motor disabilities. In this study, we introduce an innovative and optimized hybrid framework designed for the robust classification of MI-EEG signals. Our approach combines the power of a Deep Convolutional Neural Network (DCRNN) with the efficiency of the Ant Lion Optimization (ALO) algorithm. This framework consists of four key phases: data acquisition, pre-processing, feature engineering, and classification. To enhance the signal quality, our work incorporates adaptive filtering and independent component analysis (ICA) during the pre-processing phase. Feature extraction is carried out using a deep autoencoder. For classification, we employ the DCRNN, and further enhance its performance with the ALO algorithm to optimize training and classification processes. The study is implemented in MATLAB and evaluated using the PhysioNet dataset. Experimental results demonstrate the effectiveness of our proposed method, achieving an impressive accuracy of 99.32%, a precision of 99.41%, a recall of 99.29%, and an f-measure of 99.32%. These results surpass the performance of existing classification strategies, highlighting the potential of our hybrid framework in MI-EEG signal classification for various BCI applications.