Prof. Yang CHAI, Associate Dean (Research) of the Faculty of Science and Professor of the Department of Applied Physics at The Hong Kong Polytechnic University (PolyU), has conducted groundbreaking research on sensory artificial intelligence (AI), paving the way for more energy-efficient, low-latency and memory-optimised AI systems. His advancements significantly enhance diverse applications such as mobile devices, IoT sensors and edge computing.
Overcoming crucial barriers in power consumption, latency and memory within sensory AI systems, Prof. Chai's innovations unleash the potential of sensory AI across diverse industries and domains. Furthermore, the in-sensor computing strategy has sparked progress in improving decision-making and situational awareness, strengthening privacy and security, and transforming intelligent automation.
For his exceptional innovation, Prof. Chai has been honoured as a 2024 Falling Walls Winner in the Engineering & Technology category for "Breaking the Wall of Efficient Sensory AI Systems". He has developed novel hardware architectures and optimisation techniques, which enable the deployment of advanced sensory AI systems in mobile devices, IoT sensors and edge computing, subsequently transforming applications in smart cities, autonomous vehicles and industrial automation.
The Falling Walls Science Breakthroughs of the Year Award was initiated by the Berlin-based Falling Walls Foundation, to recognise the latest breakthroughs and outstanding science projects worldwide. This year, the jury, comprising globally recognised experts in the various fields, reviewed over 1,000 entries from 52 countries. In the Engineering & Technology category, 10 excellent winners were selected and shortlisted for the award of Science Breakthrough of the Year 2024.
Prof. Chai said, "The proliferation of data from ubiquitously distributed sensors leads to a massive increase in sensory terminals. It is crucial to partially shift computation tasks to the sensory terminals. This transition substantially compresses the collected information and extracts key data, especially for sensor-rich platforms."
Prof. Chai's research clearly defines near-sensor concepts and in-sensor computing paradigms based on the physical distance between sensory and computing units. This classification further divides functions into low-level and high-level processing. His study explores the implementation of near-/in-sensor computing for different physical sensing systems and provides possible solutions for integrating sensing and processing units through advanced manufacturing technologies.
While Prof. Chai and his team focus on advancing computational hardware for sensory AI systems, the extraordinary capabilities of natural bioinspired sensory systems more broadly are a vital research inspiration.
By emulating human visual adaptability, which allows accurate object identification under various lighting conditions, the new bioinspired sensors developed by Prof. Chai's team offer a solution for the progress in motion processing by directly adapting to different light intensities. This approach avoids relying solely on backend computation, which emulates and even surpasses the human retina's ability to adapt to various lighting levels. The sensors reduce hardware complexity and boost image contrast in varied lighting conditions, thus improving machine vision systems for visual analysis and identification tasks.
Inspired by flying insects' high flicker function frequency, Prof. Chai has pioneered research on optoelectronic graded neurons for perceiving dynamic motion. This innovation efficiently encodes temporal information at sensory terminals, reducing the amount of visual data transferred relating to fusing spatiotemporal (spatial and temporal) information in a computation unit. This advances machine visual systems with minimal hardware resources, promising potential applications in autonomous vehicles and surveillance systems.
These outstanding findings have been published in high-impact journals including Nature Electronics and Nature Nanotechnology, and have been highlighted in Nature, IEEE Spectrum and others, while also being highly cited by research teams worldwide.
He envisions, "My long-term goal is to develop cutting-edge microelectronic and nanoelectronic devices with new functionalities and unprecedented performance. Specifically, we intend to create imaging technology capable of perceiving three-dimensional depth, four-dimensional spatial-temporal and multiple spectral (beyond visible light) information. To achieve this, a bio-inspired mechanism will be utilised to reduce power consumption and latency."
Learn more about Prof. Chai's research focus in the video: https://www.youtube.com/watch?v=Lk7Rga3kSoc
