An Improved Algorithm for Adaptive Streaming

* Corresponding author Abstract—With the development of network facilities and the wide spread of smart phones, traditional single bitrate video streaming strategy is no longer able to satisfy people's growing needs for high quality video streams. Most of today's adaptive streaming strategies are based on bandwidth estimation such as Smooth Streaming by Microsoft IIS and HDS by Adobe. Bandwidth estimation is less accurate in network jitters which may lead to too many unnecessary stream-switches. Some other strategies are based on client-side metrics monitoring such as dropped frames, average throughput, CPU usage, buffer level. In this paper, we present a strategy that combines bandwidth estimation and client-side metrics monitoring. The client-side controller monitors the variation of the buffer level and triggers the stream-switching event accordingly. When the level of the buffer reaches to a relatively stable state, the bandwidth- estimation event is triggered to estimate the available bandwidth. If higher bitrate stream is available with the currently estimated bandwidth provided, then the stream-switching event is triggered to switch to a higher bitrate stream.