IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC)
Signal Denoising for Rydberg Atomic Quantum Receiver: A Learning-Based Duality Framework
Shengyun Liu, Jian Xiao, Zhimeng Qi, Ji Wang, Hongbo Xu, Xingwang Li, Wenwu Xie
Abstract
Rydberg atomic quantum receivers (RAQRs) offer unprecedented sensitivity for quantum sensors. However, their practical deployment is severely compromised by the extrinsic environmental noise, quantum projection noise, photon shot noise and electronic thermal noise. To address this, we propose a deep learning-based end-to-end signal denoising framework called bidirectional state space duality (Bi-SSD). The framework integrates SSD blocks to capture the time-frequency features of intermediate-frequency (IF) signals measured by RAQRs while maintaining linear computational efficiency. The bidirectional processing enables the model to utilize both past and future contextual information, thereby enhancing signal reconstruction accuracy. Furthermore, we propose an overlap-add sliding window inference strategy combined with signal statistical normalization, which enables artifact-free processing across dynamically varying sequence lengths. Evaluations on measured data of RAQRs demonstrate that Bi-SSD achieves 96.6% reconstruction accuracy under a strict 1.0% error tolerance, and delivers an $8 \times$ inference speedup over conventional attention-based models on 5000-step continuous sequences.