Wireless video multicast is increasingly popular, driven by users' desires to watch live TV and sporting events on laptops and handheld devices. The problem however is that wireless users have very different channel characteristics. Transmitting the video separately to each user is unscalable, yet transmitting a single video stream that serves all users despite their diverse channels is a challenging problem.
This project introduces Digital Rain, a new paradigm for streaming video to heterogeneous wireless receivers. Digital Rain builds on the theory of compressed sensing to develop a novel approach to video multicast where the source simply broadcasts coded packets belonging to a video stream. Receivers exploit all received video packets, even those with bit errors. Receivers that collect more packets, or experience fewer bit errors in each packet, see higher video quality without any need to receiver feedback, retransmissions, or changing the rate of the codec.
This research creates a bridge between the theory of compressed sensing and practical wireless network design. Further, it significantly improves the scalability of wireless video multicast, and hence has a direct and broad impact on emerging wireless video applications (e.g., IPTV, or Television over Internet Protocol), as well as the corresponding wireless industry. The research is integrated and synergistic with an educational plan that, in addition to standard graduate and undergraduate research and course development, directly supports MIT's effort to provide a more nurturing environment for women and underrepresented minorities. Finally, as in previous funded work, the project will result in open source, downloadable code.
