Cao Ruijie, Guo Gepu, Yue Wei, Huang Yang, Li Xinpeng, Kai Chengzhi, Li Yuzhi, Tu Juan, Zhang Dong, Xi Peng, Ma Qingyu
School of Computer and Electronic Information, Nanjing Normal University, Nanjing 210023, China.
Department of Biomedical Engineering, College of Future Technology, Peking University, Beijing 100871, China.
Research (Wash D C). 2023 Dec 1;6:0280. doi: 10.34133/research.0280. eCollection 2023.
With unlimited topological modes in mathematics, the fractional orbital angular momentum (FOAM) demonstrates the potential to infinitely increase the channel capacity in acoustic-vortex (AV) communications. However, the accuracy and stability of FOAM recognition are still limited by the nonorthogonality and poor anti-interference of fractional AV beams. The popular machine learning, widely used in optics based on large datasets of images, does not work in acoustics because of the huge engineering of the 2-dimensional point-by-point measurement. Here, we report a strategy of phase-dislocation-mediated high-dimensional fractional AV communication based on pair-FOAM multiplexing, circular sparse sampling, and machine learning. The unique phase dislocation corresponding to the topological charge provides important physical guidance to recognize FOAMs and reduce sampling points from theory to practice. A straightforward convolutional neural network considering turbulence and misalignment is further constructed to achieve the stable and accurate communication without involving experimental data. We experimentally present that the 32-point dual-ring sampling can realize the 10-bit information transmission in a limited topological charge scope from ±0.6 to ±2.4 with the FOAM resolution of 0.2, which greatly reduce the divergence in AV communications. The infinitely expanded channel capacity is further verified by the improved FOAM resolution of 0.025. Compared with other milestone works, our strategy reaches 3-fold OAM utilization, 4-fold information level, and 5-fold OAM resolution. Because of the extra advantages of high dimension, high speed, and low divergence, this technology may shed light on the next-generation AV communication.
在数学中具有无限拓扑模式的分数轨道角动量（FOAM）显示出在声涡（AV）通信中无限增加信道容量的潜力。然而，FOAM识别的准确性和稳定性仍然受到分数阶AV光束的非正交性和抗干扰性差的限制。广泛应用于基于大量图像数据集的光学领域的流行机器学习，由于二维逐点测量的巨大工程工作量，在声学领域并不适用。在此，我们报告一种基于成对FOAM复用、圆形稀疏采样和机器学习的相位错位介导的高维分数阶AV通信策略。与拓扑电荷相对应的独特相位错位为识别FOAM并从理论到实践减少采样点提供了重要的物理指导。进一步构建了一个考虑湍流和失准的简单卷积神经网络，以实现稳定且准确的通信，而无需涉及实验数据。我们通过实验表明，32点双环采样可以在拓扑电荷范围从±0.6到±2.4、FOAM分辨率为0.2的情况下实现10位信息传输，这大大减少了AV通信中的差异。通过将FOAM分辨率提高到0.025，进一步验证了无限扩展的信道容量。与其他具有里程碑意义的工作相比，我们的策略实现了3倍的轨道角动量（OAM）利用率、4倍的信息水平和5倍的OAM分辨率。由于具有高维度、高速度和低差异等额外优势，该技术可能为下一代AV通信带来曙光。
