Blind Residual CFO Estimation via CNN-Enabled EM Algorithm

• Published in: IEEE Vehicular Technology Conference pp. 1 - 5
• Main Authors: Cai, Penghao, Yang, Fuqian, Xue, Zhipeng, Wang, Yutao, Zhu, Hanyu, Gao, Xiqi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 24.06.2024
• Subjects: blind CFO estimation; CNN; EM algorithm; Estimation; Frequency modulation; Neural networks; Numerical simulation; OFDM; residue carrier frequency offset (CFO); synchronization; Vehicular and wireless technologies; Wireless communication
• ISSN: 2577-2465
• DOI: 10.1109/VTC2024-Spring62846.2024.10683526

Large residual carrier frequency offset (CFO) can severely degrade the performance of orthogonal frequency division multiplexing (OFDM) wireless communication systems when high-order modulations are adopted. In this paper, we propose a convolution neural network (CNN) enabled expectation-maximization (EM) algorithm which can blindly estimate residual CFO without extra pilots. Specifically, we first show that the effects of the residual CFO can be depicted by the phase shift existing in the equalized signal. Based on this model, we design a simple CNN to get a rough estimate of the phase shift. The output of the CNN is further used to initialize an EM algorithm. With this fine initialization, the EM algorithm can iteratively seek better estimates of the phase shift induced by the residual CFO. The combination of CNN and EM algorithm simplifies neural network design while maintaining the accuracy of the estimation. Numerical simulations verify the efficiency of the proposed method.