Human Interactive System for Networked Video Streaming
Yunlong Feng, ユンロン フェン, Yunlong Feng · Institutional Repositories DataBase (IRDB) · 2014
In this thesis, we discussed the application and development using human interactive technology and visual saliency map analysis for the networked video streaming. As we know, for current video streaming systems, there are more and more devices to record videos, more and more powerful servers to process encoding, larger and larger databases to store them, and also faster and faster network speed to transmit video packages to viewers. However, viewers still feel discontented on passively transmitted videos, for preferring a real time interactive system. In this thesis, we developed a Human Interactive System to make network video streaming more efficient, not only to benefit the network traffic, but also to adapt video content based on viewers' interests and real-time requirement. For interactive media applications, eye gaze is now used as a content adaptation trigger, such as customized advertisement in video, and bit allocation in streaming video based on region-of-interest (ROI). The reaction time of a gaze-based networked system, however, is lower-bounded by the network round trip time (RTT). Further-