To improve the performance of carrier frequency offset estimation in SC-FDE system

a nonlinear state space model is built and a new algorithm based on Unscented Gaussian particle filtering(UGPF) is proposed. In the time update step

the algorithm approximates the filtering and predictive densities of the state variables by Gaussian distributions in a PF framework

which make the substituted distribution approximating the true posterior distribution more accurately. The UKF is introduced into the algorithm to use the latest observation for particle samples modification in the measurement update step

which avoids the particle impoverishment problem because the resampling process is not need. Simulation results are presented to corroborate that compared with the existing approaches based on ML

SPF and GPF

the new algorithm not only improves the estimation accuracy and convergence speed

but also has a good performance in non-Gaussian noise environment.