Improved maximum likelihood frequency offset estimation based on likelihood metric design

For emerging high data-rate communication systems in highly dispersive channels such as ultrawideband (UWB) systems, possible frequency offsets could be larger than the estimation range of the existing methods using training signals with identical parts or repetitive training signals (i.e., the training signals are composed of several identical subblocks or are obtained by repeating a training subblock for several times). This paper presents a novel improved maximum likelihood frequency offset estimator which can handle at least twice the estimation range of the existing methods using training signals with identical parts and achieves a better estimation performance. Based on the likelihood metric, a new design metric is introduced which is a pair-wise error probability (PEP) between the correct frequency offset point and a trial frequency offset point. The proposed PEP metric gives more theoretical insights on the performance of practical maximum likelihood estimators. How to design the PEP to achieve both a larger estimation range and a better estimation performance in fading channel environments is also presented and the corresponding estimator implementation is described.