Neuromorphic Computing Revolutionizes Eye-Tracking with Ultra-Low Power

A new approach to eye-tracking technology is being developed using neuromorphic computing. Researchers have redesigned spiking neural networks to improve accuracy while drastically cutting power use and latency. The goal is to achieve real-time tracking of even the fastest eye movements in wearable devices.

Traditional eye-tracking systems often face challenges with motion blur and high energy demands. These issues limit their effectiveness in real-time applications, particularly for augmented and virtual reality. To address this, scientists are turning to neuromorphic computing, which mimics the brain's processing methods.

The team's redesigned spiking neural networks now achieve a mean gaze estimation error of just 3.7 to 4.1 pixels. This level of precision rivals existing high-end systems like Retina, but with a model size 20 times smaller. Theoretical power consumption has also been slashed by 850 times, dropping to between 3.9 and 4.9 milliwatts.

At a sampling rate of 1 kilohertz, the networks operate with a latency of only 3 milliseconds. This efficiency brings the technology closer to tracking rapid eye movements at kilohertz speeds. The breakthrough relies on replacing heavy components in current models with lightweight, energy-saving alternatives.

Event cameras, which detect only changes in brightness, play a key role in this development. Their high temporal resolution and minimal delay make them well-suited for real-time, high-precision tracking tasks.

The new system combines accuracy, speed, and low power use, making it a strong candidate for next-generation wearable tech. With further refinement, it could enable seamless eye-tracking in AR and VR applications. The reduced model size and energy demands also open doors for integration into smaller, more efficient devices.