UW professor develops terahertz sensor that processes images on-chip for wearable devices

University of Washington assistant professor Hossein Naghavi has received a U.S. Department of Energy grant to develop compact terahertz sensors that could enable wearable devices to image through materials such as smoke, clothing, and fog. The grant is part of the DOE's Genesis Mission initiative, which applies artificial intelligence across research fields.

Terahertz waves occupy the spectrum between microwaves and infrared light, combining the ability to penetrate many opaque materials with wavelengths short enough to provide useful image resolution. Conventional terahertz imagers collect raw signals at the sensor and send them to a separate processor for alignment and image formation, a process that consumes significant power and introduces latency.

Naghavi's approach integrates analog memory cells and an AI supervision layer directly into the sensor chip. This neuromorphic design aligns signals as they are collected, so the output is already organized and minimal raw data leaves the chip. The architecture is inspired by the human retina, which processes visual information locally before transmitting a compact representation to the brain.

The project is a collaboration between Naghavi's group at UW, Milad Koohi's group at Texas A&M University for hardware design, and researchers at the University of Utah and ChipNexus for AI system implementation. The immediate goal is to demonstrate a terahertz neuromorphic imager as a proof of concept in Phase I of the Genesis Mission project.

If successful, the technology could be used in augmented reality headsets for firefighting, autonomous navigation, security screening, industrial inspection, biomedical sensing, and future 5G and 6G communication networks. The team hopes to expand to Phase II to add capabilities and accelerate public accessibility.

Naghavi notes that low-cost, low-power terahertz electronics suitable for all-day wearable use have not yet been developed, and the current work aims to close that gap. The research remains in early stages, with the proof-of-concept demonstration as the next major milestone.