Resilient Video Stream Switching for Mobile Wireless Channels

In this paper, several error-resiliency techniques are combined with the H.264/AVC (Advanced Video Coding) codec’s switching frames to adaptively switch between video streams, depending on wireless channel conditions. Switching frames are a feature of H.264/AVC that allows smooth transitions between streams without the overhead of periodic intra-coded I-frames. Because video streaming over bandwidth-limited mobile networks requires higher compression ratios, when ‘lossy’ channel conditions occur it is advisable to provide error resiliency at the application layer in order to avoid degradation of video quality at the receiver. In the scheme introduced in this paper, when a change in channel conditions occurs, protection to the video stream is provided by switching to an alternative stream with error resiliency protection. To accomplish stream switching to an error-resilient video stream, minimal feedback is necessary. In this way, robust streaming is confined to periods of poor channel quality, which results in up to 3-4 dB increase in video quality (PSNR) compared to using a single robust scheme over the entire session irrespective of channel conditions. In particular, this scheme appears suited to conditions of slow fading, caused by changes in the environment as a mobile device user moves from one location to another. In the paper, the response of different error resiliency techniques to error patterns (isolated or ‘bursty’) is determined in order to adapt in a suitable way to channel conditions. Constant Bit Rate switching frames are also implemented in this paper. Compared to the usual Variable Bit Rate switching frames, CBR frames allow low-latency and low-bandwidth video services to be supported by H.264/AVC switching frames. The robust switching stream scheme can be potentially combined with adaptive stream bitrate switching. However, the main gain at low bitrates comes not from the reduced overhead of embedded switching frames but the increase in error robustness.