Adaptive rate algorithm for DASH streaming media based on probabilistic prediction
Yong Li, Bin Wang · 2017
The dynamic adaptive streaming over HTTP has been widely used in network video service. Complex network conditions require video services to have good adaptive rate. In order to improve the frequent change of client downloading bit rate and the inconvenience of network condition estimation, a new adaptive rate algorithm for DASH streaming media based on probability prediction is proposed. The sliding window is used to record the number of video slices in the buffer, and the buffer overflow probability is calculated. The overflow probability reflects the usage of the buffer. The overflow probability also indirectly estimates the network condition. In the case that the buffer in a stable range, the adjustment level of the code rate is optimized, and finally the optimal code rate is selected for downloading. Experimental results show that the rate decision algorithm can reduce the number of code rate changes and reduce the difficulty of predicting the network condition compared with the algorithm based on buffer level.