Low-Power Wireless Video Streaming and Applications

Abstract

Wireless video streaming has recently become an exciting topic of research in IoT, computer vision, and security systems. However, for some of these applications, power consumption of wireless video camera systems is a bottleneck. The issue can be addressed in two separate sections: 1- optimizing the power for wireless data transfer; since the data size in video-related applications is typically larger compared to other sensors, and 2- sensing power optimization, wherein we aim to improve video resolution while keeping the power consumption of the system low. In the first part, we focus on communication and aim to design a reliable and power-efficient wireless network of low-resolution wireless image sensors using backscatter communication technology. Also, we show that our network of low-resolution and solar-powered cameras can be used in security applications and human occupancy detection systems in buildings. Once we show that our wireless network is power-efficient, we will improve the video resolution. To enable low-power high-resolution video capturing, we use a dual-mode camera system, which switches between low-resolution gray-scale and high-resolution color video recording modes. We stream low-resolution image sensor data (gray-scale) with occasional high-resolution reference frames to the basestation. Our machine learning model implemented at the basestation generates high-resolution color frames from the low-resolution video data based on the reference frames.