Suboptimal Maximum Likelihood Frequency Offset Estimation in Flat Fading Channels

Pilot symbol assisted modulation (PSAM) is a practical technique to maintain high performance in mobile communications. Based on received pilot symbols, frequency offset can be estimated and compensated accordingly. One important class of estimators is based on the maximum likelihood (ML) criterion. The ML-based estimator has high computational complexity, which is typically of the order of O(N²). Hence, suboptimal ML-based estimators with lesser complexity are sought. In this paper, we propose two suboptimal ML-based frequency offset estimators with reduced complexity operating over flat-fading channels. The performance results of the suboptimal estimators differ from that of the ML estimator by one order of magnitude, but the computational complexity is significantly reduced for long observation lengths. The estimators are based on fast Fourier transform (FFT) and can easily be implemented on DSP architectures