PI: Yang Cai
Member: Guillaume Milcent
The bottleneck of today’s video surveillance systems is that we have too much information but not enough network bandwidth, attention and intelligence. Over 99% of data fed into the control room are wasted. The goals of this project seek to minimize the wireless video network throughput. We want to maximize the video quality with multiple resolutions on demand. In addition, this would enable the detection of events or features in the video.
We apply real-time eye gaze tracking technology to detect where the user looks at and switch the resolution accordingly.
From our empirical experiments, we found that multiple resolution screen switching can reduce the network traffic about 39%. Manual switching the 4-screens can reduce the data flow about 57%. With eye gazing interface, we obtain about 75% reduction of the network traffic.
Yang Cai and Guillaume Milcent, Attention-Aware Video Network Throughput Optimization, invited by Journal of Fuzzy Logic and Control, under view.