Research On Channel Estimation For OFDM Time-Varying Channel Based On Particle Filtering

By virtue of the good anti-multipath capability and the high spectrum efficiency, orthogonal frequency division multiplexing (OFDM) attracted widespread attention in recent years. In OFDM system, the received symbols will produce distortion, because of the time dispersion characteristics and time variation of the wireless channel. In order to recover the transmission signal accurately, we need to estimate channel. In numerous channel estimation methods, the semi-blind time-varying channel estimation method with the particle filtering has become widely accepted with its broad applicability and good performance.Starting from the particle filtering algorithm, the thesis has researched the method of time-varying channel estimation based on the particle filtering in OFDM system. The main contents are as follows:1. The basic idea of particle filtering and the algorithm flow are explained from the bayesian filtering. And the selection of the importance function and resampling techniques are discussed in detail, we focus on three improved algorithms—EPF,UPF and PF-MCMC, and compare performances of several filtering algorithms in state estimation by simulation experiment. Simulation results show that, on the basis of increasing proper complexity, the UPF algorithm can achieve better estimation performance.2. Introducing the UPF algorithm in the time-varying channel estimation field, A time-varying channel estimation method based on the UPF algorithm is proposed in OFDM system. By modeling the time-varying channel for first-order AR model with certain parameter, channel states are estimated dynamically from the observed equation by using the UPF algorithm. Simulations are carried out and simulation results show that the method can acquire higher estimation accuracy and better system performance than the method with the PF algorithm under the gaussian and non-gaussian noise environment.3. A joint estimation method based on the particle filter about the model parameter and time-varying channel states is proposed. In this method, the coefficient of the AR model is assumed to be unknown dynamic parameter, and it realizes the joint estimation combining with the kernel smoothing technique and particle filtering algorithm. Simulations are carried out and simulation results show that, the method has more significantly improvement in estimation precision and system performance relative to the channel estimation method with the PF algorithm under AR model of certain parameter; Moreover, it has greatly reduced computing complexity compared with the channel estimation method based on the UPF algorithm with AR model of certain parameter under the close performance.