Realization of High Resolution Human-Machine Interfaces by Advanced Materials
Video
Team Information
Team Members
Han Yu, PhD Candidate, Department of Electrical Engineering, Columbia Engineering
Faculty Advisor: Dion Khodagholy, Associate Professor, Department of Electrical Engineering, Columbia Engineering
Abstract
The ability to decode hand gestures is critical for natural human-machine interaction, immersive virtual reality, and creation of effective prosthetics. Although technologies such as video tracking have been applied for hand gesture recognition, they restrict the user’s mobility as they need to capture images of the body with several optical devices surrounding the user, and the cameras must maintain clear line-of-sight with all movements. Therefore, bioelectronics that are able to directly acquire and communicate the electrical activity of the human peripheral nervous system to machines have the potential to overcome these limitations. However, it is challenging to non-invasively acquire high-resolution electrophysiology signals that allow representation of ongoing muscle activity of the body. Effective contact and adhesion between conformable high-density electrodes and human skin are essential for the success of this kind of interface. Here we present a novel device that combines conformable electronics and organic mixed-conducting particulate composites (MCPs) to acquire reliable and high-spatiotemporal resolution muscle activity at the level of individual motor neuron action potentials. We demonstrate that MCP can establish isotropic conduction with the tissue allowing creation of high-density conformable electronics for the peripheral nervous system. We developed a 10x12 high density conformable array of electrodes combined with mixed conducting composites with similar particle size as skin roughness. We performed electrophysiological recordings in humans and non-human primates while subjects performed predefined tasks and movements. We were able to cluster the acquired action potentials into a large number of putative single motor units (MUs) using a template matching-based clustering algorithm. The action potential firing pattern was then correlated with the corresponding movement to establish a decoding model to allow construction of gestures. As such, our approach facilitates investigation into the neural mechanics of our sensorimotor system and enables creation of effective bioelectronic interfaces.
Team Lead Contact
Han Yu: hy2562@columbia.edu
