tinyML Talks: MicroCam: A Low-Power and Privacy Preserving Multi-modal Sensor Platform for Occupancy

MicroCam: A Low-Power and Privacy Preserving Multi-modal Sensor Platform for Occupancy Detection
Senem VELIPASALAR, Professor
Syracuse University
Heating, ventilation, and air conditioning (HVAC) consumes a significant portion of the energy used in buildings. Much of this is wasted energy, used when buildings are either not occupied at all, or occupied well under their maximum design conditions. In this talk, we will focus on residential occupancy detection to autonomously control HVAC systems and save energy. After discussing the limitations of existing solutions, MicroCam will be introduced. MicroCam is a low-cost, high accuracy, standalone residential occupancy sensing platform, which can operate on typical alkaline batteries without relying on the cloud or external computing resources, and consists of low-power, Artificial Intelligence (AI)-based, IoT platforms. Each platform has multi-modal sensors and can process motion, audio and video data. All sensor data is processed locally on platforms, and the only transmitted data is the binary occupancy state. MicroCam has been evaluated extensively, and the lessons learned will be presented.