Brain machine interfaces: recognition of mental tasks using neural networks and PSO learning algorithms / Hema C.R. ...[et al.]

• Main Authors: C.R., Hema, M.P., Paulraj, Yaacob, S., Adom, A.H., R., Nagarajan
• Format: Article
• Language: English
• Published: UiTM Press 2009
• Subjects: Pattern recognition systems
• Online Access: https://ir.uitm.edu.my/id/eprint/61860/1/61860.pdf; https://ir.uitm.edu.my/id/eprint/61860/; https://jeesr.uitm.edu.my/v1/

Brain machine interface (BMI) provides a digital channel for communication in the absence of the biological channels. BMIs are used to rehabilitate patients with neurodegenerative diseases, a condition in which all motor movements are impaired including speech leaving the patients totally locked-in. BMIs are designed using the electrical activity of the brain detected by scalp EEG electrodes. Classification of EEG signals extracted during mental tasks is a technique for designing a BMI. In this paper five different mental tasks from two subjects were studied, combinations of two tasks are studied for each subject. Two neural network architectures using a novel particle swarm optimization (PSO) learning algorithm is studied. Band power features of the EEG signals are used for the classification. The classification performance of the functional link network is seen to be higher than an Elman network. Baseline and Math tasks were found to be more suitable in designing the BMI. The results obtained validate the performance of the PSONN algorithm for mental task classification.