Non Linear Optimised Spatial Filter for Single-Trial Motor Related Cortical Potential Identification

The Motor Related Cortical Potential is a low frequency negative shift in the EEG signal appearing around 2 seconds before a planned or executed voluntary movement. Its detection is instrumental in the development of Brain Computer Interfaces which allow patients who are otherwise unable to do so to comunicate, as well as in neurorehabilitation of patients with motor impairments. An improvement in accuracy of the detectors could cause a significant advancement in the field of neuroprosthetics. In this thesis we develop a new technique to identify these potentials from EEG recordings of motor execution by calculating the optimal non linear combination of channels which isolates the signal we require. The method shows promising results, fast execution and requires a small amount of training examples, enabling training on a patient by patient basis to take into account the natural individual variability in how MRCPs manifest themselves, and making the system agnostic to the number of channels and placement of electrodes.