Design And Analysis Of Unequal Error Protection Spinal Codes

In many practical applications,some portion of data may need more protection than the other parts.For example,in wireless network,control signals may potentially be better protected than the payload data,which implies the unequal error protection(UEP)property.While in some other applications,some portion of data may need to be recovered prior to the rest of data,such as the video-on-demand system,where data should be recovered in sequence according to the requirements of different users.This corresponds to the unequal recovery time(URT)property.These applications raise a need for designing codes with UEP and URT properties.This dissertation is on the design and performance analysis of rateless UEP code,while can provide UEP and URT properties,based on rateless spinal code.The main work of this dissertation includes the following aspects:A serial UEP spinal code is proposed for binary symmetry channel(BSC)and additive white Gaussian noise(AWGN)channel.The serial UEP spinal code achieves UEP property by two methods.One is the serial encoding of message bits with different priority levels,which means that the message bits with higher priority level are encoded first,and then the message bits with lower priority level.The other is the setup of different segment sizes for message bits with different priority levels,which means the message bits of higher priority level with higher rate and message bits of lower priority level with lower rate.In addition,an unequal length transmission scheme is proposed to improve the transmission rate by avoiding the transmission of redundant encoded symbols.Simulation results show that the serial UEP spinal code can provide close-to-capacity transmission rate even at short message length,and can also offer UEP and URT properties.Performance analysis is presented for the serial UEP spinal code,including finite-length performance analysis and asymptotic performance analysis.For the finite-length analysis,upper bound on the error probability is derived for each priority level based on the potential pairwise independent property of the serial UEP spinal code.In addition,the discussion on the optimal design of the serial UEP spinal code in practical scenarios is provided by utilizing the derived upper bounds.Numerical evidence shows the efficiency of the derived upper bounds,which implies that the derived upper bounds can well evaluate the error performance of each priority level.For the asymptotic analysis,this dissertation first shows the existence of a serial UEP spinal code,whose transmission rate can be arbitrarily close to the capacity,by a lemma.Then,by employing the sliding window method,this dissertation proves that the serial UEP spinal code with a pruning decoder can achieve the capacities of both BSC and AWGN channel.A superposition UEP spinal code is proposed for the binary erasure channel(BEC).By the superposition operation,the information of message bits with higher priority level is conveyed by more coded symbols than that with lower priority level,which results in UEP property.And different UEP performance can be flexibly achieved to meet different scenario requirements by adjusting the superposition parameter.In addition,two decoding strategies are proposed based on the encoding structure of the superposition UEP spinal code,and an interrupting decoding parameter is introduced to decrease the decoding complexity.Simulation results show that the superposition UEP spinal code has good error-correcting performance while providing UEP and URT properties,and different UEP performance can be achieved by employing different superposition parameters.Finite-length performance analysis is provided for the superposition UEP spinal code.During the finite-length performance analysis of superposition UEP spinal code,the probability model of received signal is first discussed in a lemma.Then,upper bound on each priority level is formulated in a theorem,and the corresponding proof is also presented in detail.Finally,this dissertation compares the error probability of each priority level and the derived upper bounds.It can be observed that the derived upper bound is very close to the error probability of each priority level by practical simulations,which indicates that the derived upper bounds can accurately estimate the error probability of each priority level.