Low-complexity Channel Estimatin For Cooperative Wireless Sensor Networks With Data Censoring

Wireless sensor networks(WSNs)are developing rapidly in recent years since the wide range of applications.One of the most important problem on the WSNs is the consumption requirement in communication resources.In some situations where sensor nodes send large scale data to each other,data censoring scheme can be utilized to save communication overhead.In this case,low complexity signal processing algorithms with data censoring schemes are developed.In this dissertation,we focus on the problem of channel estimation for a general wireless sensor network(WSN)with data censoring.In this case,a cooperative WSN system with data censoring is considered.Two data censoring schemes which include non-adaptive censoring(NAC)scheme and adaptive censoring(AC)scheme are presented for reducing the transmission consumption.Then,based on the WSN system model with NAC and AC,channel estimation algorithms are proposed.For NAC scheme,the stochastic gradient descent(SGD)and Newton method are employed to find the maximum likelihood estimator(MLE),respectively.The main objective of these two algorithms is to reduce the computational complexity of MLE.Furthermore,the lifetime of WSN is extended with NAC scheme by saving transmission resources.For AC scheme,the least mean square(LMS)and recursive least square(RLS)are applied to estimate the unknown channel.The set-membership(SM)framework is applied in AC to get a further improve the convergence performance.Moreover,the computational complexity is reduced with AC scheme by data selection.In addition,the censoring ratio(CR)analyses of NAC and AC schemes are presented for achieving a desirable tradeoff between CR and the convergence performance.The simulation results show the good performance of the proposed channel estimation algorithms which achieve similar steady state mean square error and convergence speed with less transmitted data and lower computational complexity than conventional channel estimation algorithms without data censoring.