A Novel Principal and Independent Component Analysis Preprocessing Technique for Neural Network Classification of Electroencephalography Signals for Brain Computer Interface Development

摘要

The field of brain computer interfaces (BCI) is growing rapidly. Innovations that help benefit disabled persons is the overall goal of the research, currently. Every brain computer interface consists of three basic parts: a sensing device, signal processing, and an actuator. This work contributes to the second of the three parts, signal processing. This work presents and tests a novel method for combining the already established work of principal component analysis, independent component analysis, and artificial neural networks to generate a brain computer interface for controlling a robotic hand. The sensing device is substituted by a data set and the actuator is substituted by using a simulator. This work also presents a framework for rapid development using this method and testing inside the simulated environment with different hardware to ease the transition from the theoretical to the practical.;Results of the developed algorithm were assessed with current state of the art techniques and was found to be competitive or more robust than other techniques. The algorithm was evaluated across 10 subjects, with typical results from one subject presented. Imagined left and right hand grasp intent were classified, along with another classifier for neither intent.