Robustness of joint Bayesian frequency offset and channel estimation based on basis expansion models

This paper deals with data aided joint Bayesian frequency offset and frequency-flat time-variant channel estimation based on basis expansion models (BEM)s of the time variation. The approach exploits the knowledge of channel statistics, in particular, the Doppler frequency. We investigate the sensitivity of the estimator using such BEMs as Karhunen-Loe¿ve (KL), discrete prolate spheroidal (DPS), generalized complex exponential (GCE), and B-spline (BS) functions to the knowledge of the Doppler frequency. Simulation results show that for a perfectly known Doppler frequency, all the BEMs can achieve the same performance. However, the BS and GCE BEMs are more robust in the case of mismatched Doppler frequency.