Private:mobile streaming ideas

Mobile TV

• Extend power awareness to WiMax and TDMA networks.

• Equalizing perceived-quality for multiple concurrent TV channels: We consider a mobile TV system that broadcasts several video sequences at the same time, while these video sequences have heterogeneous characteristics/complexities and require diverse bit rates to achieve the same perceived-quality. Our goal is to determine the best streaming bit rate for each TV channel such that all TV channels achieve a uniform perceived-quality that is less than or equal to a given target quality chosen by network operators.
  • We then show that by tolerating a small quality variation (or quality degradation), we can increase the energy saving (or battery life).
  • The same solution can be applied to an Internet on-demand video server to reduce the server-load and bandwidth consumption by providing users good, but not too good, perceived quality. The target quality can be in server level agreements.

• Dependency-aware burst scheduling: Video refresh time, caused by decoding dependency among video frames, is one of the dominating sources of channel switching delay. Therefore, reducing refresh time is important for user experience because most users tend to switch channels frequently. Our goal is to reduce video refresh time without compromising bandwidth utilization. Most previous works insert/replace an immediately decode-able (IDR) frame at the beginning of each burst. Although doing so will reduce video refresh time, it also increases the bandwidth consumption because IDR frames are much larger than predictive frames. We believe this problem can be better solved by dynamically varying burst sizes according to given frame structures so that IDR frames are as close to the beginning of individual bursts as possible. For example, having an IDR frame as the 2nd frame in a burst only incurs negligible delay compared to inserting an IDR frame as the 1st frame. However, the former case consumes lower bandwidth because P- or B-frames are much smaller than I-frames.

• Statistical multiplexing: We consider VBR-coded video streams and apply statistical multiplexing method in the literature to mobile TV networks to solve burst scheduling problem. We analytically analyze the performance of such mobile TV networks in terms of energy saving, channel switching delay, packet loss rate/re-buffering instances, among others. Our analysis is based on existing models of VBR traffic flow and of Internet router design. At high level, our burst scheduler allocates bursts to individual TV channel with proportional length to their buffer levels: more air time is assigned to TV channels with more pending packets in the outgoing queues. The ultimate goal of our analysis is to design an admission algorithm that provides QoS guarantees in statistical fashion.