Coarse Frequency Offset Estimation in MIMO Systems Using Neural Networks: A Solution with Higher Compatibility

Abstract

Carrier frequency offset (CFO), which often occurs due to the mismatch between the local oscillators in transmitter and receiver, limits the performance of multiple-input multiple-output (MIMO) wireless communication systems. To recover the CFO, the first step is coarse CFO estimation. This paper presents a neural network (NN) based coarse CFO estimator which has higher compatibility with a variety of MIMO systems, comparing with traditional CFO estimators. Instead of performing closed form calculation as some traditional estimators do, the proposed estimator transforms the estimation problem to a classification problem: classify the optimal coarse CFO estimate from a pool of coarse CFO candidates. Taking the advantage of neural networks, the proposed NN estimator can perform coarse CFO estimations for MIMO systems with different numbers of antennas and a variety of channel models. Meanwhile, the testing results show that the proposed NN estimator has promising performance and wide CFO acquisition range.