tinyML Talks: ML using micro-electromechanical system (MEMS)

“ML using micro-electromechanical system (MEMS)”

Fadi Alsaleem, Ph.D.,
Assistant Professor
Durham School of Architectural Engineering and Construction
Mechanical Engineering Department
University of Nebraska – Lincoln

This talk covers a new technology that enables a sensor such as a wearable accelerometer to provide high-level processed information such as step counts or type of activity rather than the simple raw acceleration measurement. This new technology is based on a micro-electromechanical system (MEMS) where a network of them is designed to locally perform advanced algorithms. The algorithms will be coded in the mechanical responses of the sensing elements of these multiple-coupled MEMS devices that simultaneously capture the measurement of interest such as acceleration. As a result, the MEMS network will perform computing at the sensing physical layer and will require very little power, eliminating the need for a microprocessor and eliminating the need for the energy-hungry circuitry for conditioning and reading the output of the traditional sensor. The new sensing/computing technology has been demonstrated in multiple applications such as human activity recognition, simple signal classification, and mobile robot.

source