TrsNet: A TRS-based deep learning network for carrier frequency offset estimation in 5G system

This article proposes a deep learning network, TrsNet, based on Tracking Reference Signal (TRS) for carrier frequency offset (CFO) estimation in 5G systems. Due to the use of Orthogonal Frequency Division Multiplexing technology in the 5G downlink, the system is susceptible to CFO, which can lead to signal amplitude attenuation and phase distortion, thereby affecting communication performance. To address the issues above, we propose designing and implementing a deep learning CFO estimation network, TrsNet, based on TRS, which improves the accuracy and robustness of CFO estimation by learning critical features of TRS signals. Through simulation experiments under different signal-to-noise ratios and CFO conditions, the performance of TrsNet in AWGN channels was verified. The results showed that TrsNet has a strong anti-noise interference ability, which can solve the limitations of traditional algorithms in estimation accuracy and range. At the same time, compared with similar deep learning methods, the proposed network model has lower complexity and more robust adaptability. Finally, this article also explores the challenges of applying deep learning technology in 5G communication and provides prospects for future research directions.