Motor imagery based EEG signal analysis for control of mobility assistive device

Abstract

Disorders like Quadriplegia, Brain Stem Stroke, Spinal cord injury, Amyotrophic Lateral Sclerosis, Muscular dystrophy, or Multiple Sclerosis can disrupt the neural pathway by which the brain communicates and controls the limbs. In such cases, it can lead to mobility impairment. Mobility assistive devices such as wheelchairs require physical inputs from the upper body for their operation, which cannot be provided by people with motor impairment. Hence, there is a need for a mobility assistive device with its control independent of muscular intervention. Brain-computer interface (BCI), an advanced technology that records the brain activity to understand the user s intent and interprets them to control an external device, can be used as an alternative control for the mobility aid. Motor-impaired people can utilize the motor imagery (MI) Electroencephalography (EEG) signals acquired from the scalp when the person imagines moving the limbs for control. Hence, in this thesis, an efficient signal analysis system for identifying the user s intent from the MI EEG signals for the control of mobility assistive device is explored. The Common Spatial Pattern (CSP) spatial filtering algorithm is a widely used pre-processing and feature extraction algorithm that captures the discriminant MI patterns in MI-based BCI systems. The first contribution in this thesis is to enhance the performance of CSP-based MI classification by exploring the use of ERD/ERS (event-related desynchronization/event-related synchronization) information for processing. Additionally, multiclass MI EEG signals are acquired using a designed acquisition protocol and are used to analyse the proposed processing and classification algorithm. newline