Rate-Distortion Optimized Joint Source/Channel Coding of WWAN Multicast Video for a Cooperative Peer-to-Peer Collective

Because of unavoidable wireless packet losses and inapplicability of retransmission-based schemes due to the well-known negative acknowledgment implosion problem, providing high quality video multicast over wireless wide area networks (WWAN) remains difficult. Traditional joint source/channel coding (JSCC) schemes for video multicast target a chosen th-percentile WWAN user. Users with poorer reception than th-percentile user (poor users) suffer substantial channel losses, while users with better reception (rich users) have more channel coding than necessary, resulting in sub-optimal video quality. In this paper, we recast the WWAN JSCC problem in a new setting called cooperative peer-to-peer repair (CPR), where users have both WWAN and wireless local area network (WLAN) interfaces and use the latter to exchange received WWAN packets locally. Given CPR can mitigate some WWAN losses via cooperative peer exchanges, a CPR-aware JSCC scheme can now allocate more bits to source coding to minimize source quantization noise without suffering more packet losses, leading to smaller overall visual distortion. Through CPR, this quality improvement is in fact reaped by all peers in the collective, not just a targeted th-percentile user. To efficiently implement both WWAN forward error correction and WLAN CPR repairs, we propose to use network coding for this dual purpose to reduce decoding complexity and maximize packet recovery at the peers. We show that a CPR-aware JSCC scheme dramatically improves video quality: by up to 8.7 dB in peak signal-to-noise ratio for the entire peer group over JSCC scheme without CPR, and by up to 6.0 dB over a CPR-ignorant JSCC scheme with CPR.