Optimized Neural Network with Refined Features for Categorization of Motor Imaginary Signals

Abstract

Motor imaginary (MI) is an attractive research field in the brain–computer interfaces (BCIs) function, in which the system is directed by the imaginary arm movement of the subject. This attention is due to the monstrous potential for its pertinence in neurorestoration, neuroprosthetics, and gaming, where the client’s considerations of envisioned developments should be decoded. An electroencephalography (EEG) device is regularly utilized for monitoring frontal cortex movements in BCI frameworks. The EEG signals are perceived through the two fundamental processes such as feature extraction and characterization process. This research concentrates on developing a predominant MI categorization model utilizing deep learning techniques. The prominence of this research relies on the combined features + proposed PROA-based RideNN process known as holo-entropy-based WPD, which extracts the most dominant feature from the EEG signals. The extracted features enhance the performance of the RideNN classifier. The analysis is done by utilizing the BCI Competition-IV-2a, -2b, and GigaScience datasets with respect to performance parameters, such as specificity, accuracy, and sensitivity. The analysis revealed the effective performance of the proposed method with respect to the existing state-of-art methodologies.