Circuit Techniques for Optimized Recording of Neural Signals

Abstract

Bidirectional Brain Computer Interfaces (BBCIs) are an emerging technology that will provide increased quality of life for patients with various neurological disorders, and will likely someday enhance the human-computer interaction among healthy human populations. The bidirectional nature of these systems mean that many hundreds to thousands of sense and stimulation electrodes must exist simultaneously in human tissue. This complex system creates a number of engineering challenges that are explored in this work, which discusses techniques for accurately and completely recording sensed information from the brain. In this dissertation two primary efforts are discussed that utilize the character of biological signals to build optimal, robust, and efficient systems for recording data from the brain. Data-driven specifications are described for a particular type of neural signal, electrocorticography, which is recorded from electrodes on the brain surface. Based on this work, a novel mixed-signal feedback architecture is exploited to create a robust and efficient neural recording platform, including time-domain multiplexing to reduce the silicon area required for each electrode recording site, and stimulation artifact suppression, to maintain the integrity of the recorded data in the presence of bidirectional communication with the tissue. Measurement results will demonstrate the utility of these data-driven specifications and the success of this new architecture.