On delimiting video rebuffering for stream-switching adaptive applications

Most of the current adaptation algorithms assess download rate, or trace playout buffer occupancy to adapt video representation. We propose a novel approach for switching video representation based on the estimated probability of video rebuffering. This probability is calculated using an analytical model of playout buffer and measured segment download time characteristics. The results of simulation experiments and prototype trials, carried out in a controlled network environment and over the Internet, show that our method efficiently accommodates changing network conditions.