Unequal Error Protection for Video Transmission over Wireless ChannelsOutlineUEP/ULPUEPProblem FormulationSlide 6Channel ModelReed-Solomon (RS) codeAnalysisTheory results-BSCTheory results-BSC (contd.)Simulation results (1)Simulation results (2)Simulation results (3)Simulation results (4)Conclusion1Unequal Error Protection for Video Transmission over Wireless ChannelsECE738 Project PresentationChang, Hong Hong05/09/20032OutlineUnequal Error Protection/ Unequal Loss ProtectionProblem FormulationChannel ModelRS codeTheoretical ResultsSimulation ResultsConclusion3UEP/ULPVideo quality is determined by the packet loss behavior observed at the video decoder. Time varying wireless channelsTime varying networking conditionsLimited transmission rate Different priority source data4UEPGoal of UEPAssign unequal amounts of error correction /protection bitsTo gain optimal quality in the receiver sideSome of algorithmsRate Compatible Punctured Convolutional codes Multiple description Priority encoding transmission - unequal amount of FEC packets5Problem FormulationUse Reed-Solomon code as the channel codingGiven a fixed size of packet, how many protection bits/symbols can be added in the RS codeword to get the optimal quality in the receiver side?6Problem FormulationOptimization criteriaMaximize expected PSNRMinimize expected distortionMaximize expected effective source rate (used in this project)Work need to doChannel model + RS coding -> Packet loss ratePacket loss rate -> effective source rate7Channel ModelGoodBadPGBPBGTwo state Markov Channel Model GBBGB BGPPP P=+8Reed-Solomon (RS) codeThe RS (n, k) channel code converts every k information symbols into an n-symbol block by appending n-k parity symbols. Any error pattern with less than symbol errors can be corrected. nk2tdata parity2n kt-� �=� �� �9AnalysisSymbol error probability PB Probability of i symbol errors within a block of n successively transmitted symbols Packet loss rate Effective source rate for one packet k*(1-PL) Calculate expected effective source rate over several packets( , ) (1 )i n iD B BnP n i P Pi-��= -����1( , )nL Di tP P n i=+=�10Theory results-BSC50 100 150 200 250 300050100150200250kEfective source rateEfective source rate as function of k, n = 255PB =0.05PB =0.1PB =0.15PB =0.211Theory results-BSC (contd.)100 150 200 250 300 350 400 450 500 550050100150200250300350400450kEfective source rateEfective source rate as function of k, n = 511PB =0.05PB =0.1PB =0.15PB =0.212Simulation results (1)200 205 210 215 220 225185190195200205210215kEfective source rateEfective source rate as function of k, n = 255 PB = 0.05TheoreticalSimulation13Simulation results (2)170 175 180 185 190 195170172174176178180182kEfective source rateEfective source rate as function of k, n = 255 PB = 0.1TheoreticalSimulation14Simulation results (3)150 155 160 165 170 175120125130135140145150155kEfective source rateEfective source rate as function of k, n = 255 PB = 0.15TheoreticalSimulation15Simulation results (4)120 125 130 135 140 145112114116118120122124126128130kEfective source rateEfective source rate as function of k, n = 255 PB = 0.2TheoreticalSimulation16ConclusionPB is smaller -> k is larger, better optimal effective source rate.The simulation results are basically match with the theoretical
View Full Document