The objective of this research is to discover a new class of ultra low power sensors. The approach is to develop nonlinear adaptive schemes that circumvent the tradeoffs faced by linear, time-invariant systems. This research aims at facilitating applications like mobile health, which rely on sensors with a limited power budget. In particular, this research will enable the brain-controlled cochlear implant - a next generation neural prosthetic that addresses the long-standing problem of speech intelligibility in a noisy background.
Intellectual Merit: Nonlinear analysis can identify those regions of sensor data that are critical for an application's computations and state changes. Further, an adaptive system can adjust its parameters and resource consumption to specifically accommodate such computationally-relevant data, while discarding the remaining, less relevant data. These nonlinear and adaptive schemes result in very power-efficient sensors, and the proposed project will realize them by methodically harnessing the inherent nonlinear dynamics of analog components. In this way, the project aims to advance the field of low power sensor design.
Broader Impacts: The mobile health devices that this project enables will promote a preventative, decentralized model of health-care delivery. This could allay the rising cost of health care. The project will integrate nonlinear systems thinking into engineering teaching through outreach programs and interdisciplinary workshops that are already thriving at Dartmouth College. The project's health applications will demonstrate that engineering is a human-focused occupation. This is relevant to women and minorities who, it has been suggested, avoid engineering because of its perceived disconnect with direct human concerns.
