Affordable delay based quality selection for HTTP adaptive video streaming

We present a quality-selection policy for Quality of Experience (QoE) demanding video streaming in wireless networks. The proposed policy predicts the TCP throughput and adapts video segment requests in order to assure high QoE by taking into account the client buffer level. We introduce the concept of Affordable delivery Time (AT) and we design a buffer-based algorithm, hereafter referred to as Buffer-based lolyPOP (BPOP). The AT accounts for the ideal chunk download time which equals the chunk playout time, as well as for the client buffer status. In a nutshell, the buffer-based AT favors the download of higher quality video chunks when the client has more buffered data than a pre-established target buffer occupancy. Conversely, AT inhibits higher quality video chunks download when the buffer is below the targeted occupancy.