The last decade has seen a surge in the development of brain-machine interfaces (BMIs) as assistive neural devices for paralysis patients. BMIs are devices that decode neural activity to provide control signals for external devices, computers or prostheses. Current BMI research typically involves a subject playing a computer game or controlling a robotic prosthesis through neural activity (brain control). The local field potential (LFP) is a low frequency neural signal recorded from intra-cortical electrodes, and has been recognized as one containing movement information. This thesis investigates time and frequency properties of the LFP from the perspective of developing upper limb neuroprosthetic BMIs, and touches on three major topics.

First, the thesis considers LFP as a direct input for BMIs. The results in the thesis confirm previous studies that established the modulation of LFP spectral power by limb movement direction during manual control. However, in addition, it is observed that all signals from an electrode array show similar direction modulation. Basic offline movement prediction from only LFP information is also demonstrated. Second, the thesis explores coherence between two LFP signals. Results note that LFP signals across regions of the motor cortex are strongly coherent in the beta range (15-45 Hz) during stationary periods of manual control experiments. Some technical considerations for LFP-LFP coherence calculations are presented. Finally, brain control experiments with single unit action potentials controlling a computer cursor are considered. Findings in the thesis show that beta band LFP activity during brain control tasks closely resembles beta activity during tasks involving direct limb movement. A last set of results indicate that the LFP beta band power can predict the movement state of a brain controlled cursor.

In conclusion, this thesis demonstrates the utility of the LFP as a supplementary information signal to develop the next generation of BMIs.




Download Full History