Toward High Performance, Power Efficient Brain-Machine Interfaces

LOCATION: NCRC Bldg 32 Auditorium

PASSWORD: brains

Brain-machine interfaces (BMIs) are a promising solution for restoring mobility and communication to people who suffer from sensorimotor impairments, including spinal cord injury, stroke, and neurodegenerative diseases. Intracortical BMIs consist of an electrode array implanted in the brain, a signal processing pipeline to amplify and extract neural features, and a decoding algorithm to predict the user’s intentions. In this work I aim to develop high performance, real-time decoding algorithms that reduce the power consumption of the entire BMI, facilitating clinical translation toward fully implantable, wireless systems.

First, I optimize an efficient communication scheme for transmitting neural data from 1000s of wireless electrodes while staying within power limits. Second, I show how recurrent neural network decoders can accurately predict finger movements in real-time, outperforming other neural network architectures for closed-loop control. Additionally, I introduce a training regularization technique that controls the decoder’s degree of memorization of movement patterns. Finally, I show how neural network decoders can be compressed by over 100x, and how the number of active electrodes can be reduced by over 2x without performance loss, for large reductions in power and the ability to run the decoder directly on an implanted device.

CO-CHAIRS: Cindy Chestek and David Blaauw