In the realm of energy-efficient technologies, a trio of researchers from the University of Zurich—Federico Paredes-Valles, Yoshitaka Miyatani, and Kirk Y. W. Scheper—have made significant strides in developing a low-power, wearable eye-tracking system. Their work, published in the journal Nature Machine Intelligence, presents a novel approach to pupil tracking that could have practical applications in various industries, including energy.
Eye tracking is a crucial component in many applications, from augmented reality to user interface design. However, creating a robust, high-frequency tracking system with ultra-low power consumption has been a persistent challenge, particularly for wearable platforms. Traditional eye-tracking systems often rely on bulky, power-hungry components that are impractical for portable devices. The researchers sought to address this issue by leveraging event-based vision sensors and neuromorphic processing.
Event-based vision sensors offer microsecond resolution and sparse data streams, making them ideal for low-power applications. However, until now, they have lacked fully integrated, low-power processing solutions capable of real-time inference. The researchers developed a battery-powered, wearable pupil-center-tracking system that combines event-based sensing and neuromorphic processing on the commercially available Speck2f system-on-chip. This system is complemented by lightweight coordinate decoding on a low-power microcontroller.
The researchers’ solution features a novel uncertainty-quantifying spiking neural network with gated temporal decoding, optimized for strict memory and bandwidth constraints. They also developed systematic deployment mechanisms to bridge the reality gap, ensuring the system’s practicality in real-world scenarios. The system was validated on a new multi-user dataset and demonstrated through a wearable prototype with dual neuromorphic devices achieving robust binocular pupil tracking at 100 Hz with an average power consumption below 5 mW per eye.
For the energy sector, this technology could be particularly useful in applications requiring continuous, low-power monitoring of user interactions with interfaces, such as in control rooms or field operations. The ability to track eye movements with high precision and minimal power consumption could enhance the efficiency and safety of various energy-related tasks. Moreover, the integration of neuromorphic computing in wearable devices could pave the way for more advanced, energy-efficient technologies in the future.
In summary, the researchers have demonstrated that end-to-end neuromorphic computing enables practical, always-on eye tracking for next-generation energy-efficient wearable systems. Their work represents a significant step forward in the development of low-power, high-performance technologies that could have wide-ranging applications in the energy industry and beyond. The research was published in Nature Machine Intelligence, a reputable journal known for its focus on cutting-edge advancements in machine intelligence and its applications.
This article is based on research available at arXiv.