Signal processing techniques for motor control brain computer interface systems

• Main Author: Neethu Robinson
• Other Authors: Cuntai Guan
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Computer science and engineering::Computer applications::Life and medical sciences
• Online Access: http://hdl.handle.net/10356/62308
• Institution: Nanyang Technological University

Brain Computer Interface (BCI) is a communication and control system that establishes a non-muscular pathway between brain and external environment. BCIs aim to develop assistive devices for people with severe motor disabilities, using their cognitively intact brain. This has led to the development of motor control-BCIs using Electroencephalography (EEG) and functional Near Infrared Spectroscopy (fNIRS). The thesis aims to develop BCI algorithms to identify and extract neurophysiological phenomena underlying human motor control specifically, movement kinematics, execution and imagination of motor tasks using EEG and fNIRS signals. The challenge involved is the insufficient signal resolution of noninvasive EEG to identify the deeply encoded kinematic information. Further, feasibility of fNIRS to develop a subject-independent neurofeedback BCI system by effectively utilizing neuronal correlates of overt and covert motor tasks is also investigated. The EEG-BCI reported in the thesis incorporates novel signal processing tools to extract neural features that can discriminate movement parameters such as speed and direction, and can be used for classification and reconstruction of trajectory. The proposed processing tools include Wavelet Common Spatial Pattern (WCSP), spatially regularized WCSP, Signal-to-Noise Ratio (SNR) of direction dependent tuning of EEG signal and wavelet-based decoder models. The fNIRS-BCI reported in the thesis explores the overlapping neuronal activity of movement execution and imagination. The proposed strategy generates discriminative features for real time binary classification of bilateral movement tasks using subject independent adaptive classifier. The results achieved using the reported algorithms are found to be improved and statistically significant than the existing methods. The research presented in this thesis explored various aspects of movement control BCIs using EEG and fNIRS signals. The proposed techniques can contribute to the development of BCI systems with higher dimensional motor control and improved efficiency. The research suggests the further extension of the proposed BCI system to multimodal EEG-fNIRS system to explore motor activities that can help develop assistive devices for motor impaired users.