|
Getting your Trinity Audio player ready...
|
The Associate Dean (Research) at the Faculty of Science and a Professor in the Department of Applied Physics at The Hong Kong Polytechnic University (PolyU) has made significant strides in sensory artificial intelligence (AI), addressing key challenges such as power consumption, latency, and memory optimisation. His research is transforming how sensory AI systems are integrated into technologies like mobile devices, Internet of Things (IoT) sensors, and edge computing, making them more efficient and adaptable.
Sensory AI refers to systems designed to process data collected from the environment in ways that mimic human sensory functions. These systems hold the potential to revolutionise industries ranging from smart cities to autonomous vehicles and industrial automation. However, the primary barriers to deploying these technologies on a large scale have been high power requirements, slow data processing, and inefficient memory use.
Professor Chai’s groundbreaking work targets these issues by introducing a novel approach known as in-sensor computing, which shifts computation tasks closer to the sensory terminals. This method dramatically reduces energy consumption, speeds up data processing, and enhances memory efficiency.
A key innovation in Professor Chai’s research is the development of in-sensor computing, where sensory terminals handle part of the computational workload. By processing and compressing data at the sensor level, this strategy allows critical information to be extracted in real-time, reducing the amount of data that must be transferred to central processing units. This not only lowers power consumption but also enables faster decision-making and situational awareness, making the technology particularly useful for sensor-rich platforms like smart cities and autonomous vehicles.
One of the standout benefits of in-sensor computing is its ability to enhance both privacy and security. Since less data needs to be sent across networks, the risk of data breaches is reduced, and sensitive information can be processed locally.
Additionally, in-sensor computing is key to advancing intelligent automation, where machines can independently process sensory information without relying on centralised systems. This shift significantly enhances the performance of systems that depend on real-time data processing, such as industrial automation and smart infrastructure.
The team has developed novel hardware architectures and optimisation techniques that enable the deployment of sensory AI in various technologies. These innovations are expected to transform applications such as mobile devices, IoT sensors, and edge computing, where energy efficiency and real-time processing are crucial.
A notable aspect of his research is the development of bioinspired sensors that mimic the human eye’s ability to adapt to varying light conditions. These sensors automatically adjust to changes in light, enhancing image contrast without extensive backend computation. This is crucial for machine vision systems, which rely on clear visual data for tasks like object recognition and motion detection, essential in autonomous vehicles and surveillance systems.
Professor Chai’s research also explores optoelectronic graded neurons inspired by flying insects’ rapid motion perception. These neurons encode temporal data at the sensory terminals, minimising data transfers. This innovation is particularly beneficial for systems like autonomous vehicles, where real-time motion processing is essential, optimising hardware and reducing power consumption for more efficient machine vision systems.
Professor Chai’s work has been published in top-tier scientific journals, including *Nature Electronics* and *Nature Nanotechnology*, and highlighted in prestigious outlets such as *Nature* and *IEEE Spectrum*. His research is frequently cited by other teams around the world, reflecting its global impact on the field.
