Neuromorphic Vision Sensors, Processing, Algorithms
My main research goal is to develop useful neuromorphic or bioinspired analog VLSI chips, in particular high performance vision sensors for real-world applications.
Vision sensors such as the dynamic vision sensor take inspiration from the brain's use of activity-driven, event-based signaling. These sensors output activity-driven events, rather than conventional regular samples such as picture frames, and so inspire new forms of machine vision and audition. Their high dynamic range, low power consumption (especially at the system level), sparse, and quick output make them useful for many tasks such as robotics and surveillance.
My group develops these custom IC chips, the system level sensors (cameras and auditory processing modules with e.g. USB interfaces) and their applications in robotics and scientific research. Sensors are designed in conventional CMOS technologies ranging from 350nm to 90nm using mixed analog / digital and synchronous / asynchronous logic. Applications development in turn spurs the development of open-source, efficient, real-time event-based signal processing algorithms.