Phase Noise Estimation Based On Combined Message-Passing Algorithms

With the rapid development of wireless communications of next generation,the demands of the capacity of data,broadband data services and the number of user grow exponentially.To meet the increasing demand,Ultra-High Frequency(UHF)wireless communication becomes the developing trend.However,high precision UHF oscillator is difficult to implement for the limitations of the semiconductor manufacturing technology.Then,imperfect oscillators introduce phase noise to carrier signal between transmitter and receiver.In addition,to improve the transmission rate,the transmitter adopts high order modulation technology which is susceptible to phase noise.Therefore,phase noise is an important factor affecting the performance of wireless communication systems.Focusing on systems with presence of phase noise,the design for the receiver with low complexity and high performance becomes a research hotspot.In this thesis,considering the single input single output(SISO)and multiple input multiple output(MIMO)systems with phase noises,joint phase noise estimation,detection and decoding algorithms based on message passing are addressed.1.For additive white Gaussian noise(AWGN)channels where variance of Gaussian noise is unknown,a low-complexity phase noise estimation and detection algorithm is proposed.Different from the existing sum-product(SP)-based algorithms which discretize the unknown continuous variables,the proposed algorithm efficiently realizes joint phase noise estimation,noise variance estimation and decoding by combining variational message passing(VMP)and SP.Employing VMP to handle the non-Gaussian continuous variable message achieves low complexity.Simulation results demonstrate that the proposed algorithm has the same performance as that of the state-of-the-art Tikhonov receiver with much less computational complexity.2.Focusing on low-complexity detection problem in the single carrier system over multipath channels with phase noise,a doped expectation propagation(Doped EP)algorithm is presented by combining Gaussian approximated belief propagation(GABP)with belief-and expectation-propagation(BP-EP)detection algorithm.Comparing with BP-EP and GABP,the proposed Doped EP algorithm efficiently solves the iterative vibrant phenomenon in BP-EP,and obtains a significant bit error rate(BER)gain.3.For coding single carrier system over inter-symbol interference(ISI)channels with phase noise,an iterative receiver is designed based on belief propagation(BP),mean field(MF)and expectation propagation(EP)to perform joint phase noise estimation,equalization and decoding.By partitioning system factor graph,MF is exploited to handle the non-linear observation model,BP-EP is employed to deal with the linear model for the phase noise process and modulation and coding.Unlike the state-of-the-art soft input extended Kalman smoothing(Soft-in EKS)algorithm linearizing the nonlinear model by first-order Taylor expansion,we adopt second-order Taylor expansion to achieve Gaussian approximation,remaining the second order statistics of phase noise distribution.As shown by the simulation results,comparing with Soft-in EKS algorithm,the proposed algorithm obtains a significant performance gain with similar complexity.4.For phase noises caused by imperfect high frequency oscillators among different antennas in point-to-point MIMO system,a joint channel estimation,phase noise estimation and symbol detection algorithm is presented by merging generalized expectation maximization(GEM)and BP.The nonlinear likelihood function nodes and symbol variable nodes are processed in E-step of GEM while the marginal posterior probability density functions of channel parameter and phase noise are tracked in M-step using BP.The channel and phase noises are estimated by exploiting the training sequence and the detected symbols to improve the estimation precision.Numerical simulations demonstrate that the proposed algorithm outperforms the advanced soft-input extended Kalman smoothing approach in terms of mean square error,bit error rate and frame error rate with similar computational complexity.And for uncoded system,the proposed algorithm can approach the performance under the scenario with perfectly known channel state information and phase noise.