【摘要】 针对传统频谱感知在复杂信号环境，如低信噪比和多路径衰落等情况下的性能局限，提出了一种基于黎曼流形神经网络的创新性协作频谱感知方案。该方法首先通过将多个协作用户的信号矩阵映射到黎曼流形上，生成具有几何特性的协方差矩阵。接着，利用黎曼流形神经网络进行高效的信号特征分类和频谱感知。黎曼流形神经网络不仅充分发挥了黎曼流形在非欧几里得数据结构上的优势，而且结合了神经网络强大的表达能力，从而在各种复杂环境下都展示出显著优越的频谱感知性能。通过一系列详细的仿真实验，证明了该方法在多样化环境下的优越性能，展示了其在实际无线通信系统中的潜在应用价值。更多还原

【Abstract】 This study proposes an innovative collaborative spectrum sensing scheme based on Riemannian manifold neural networks, addressing the performance limitations of traditional spectrum sensing in complex signal environments, such as low signal-to-noise ratios and multipath fading. The method initially maps the signal matrices from multiple collaborative users onto a Riemannian manifold, generating covariance matrices with geometric characteristics. Subsequently, Riemannian manifold neural networks are utilized for efficient signal feature classification and spectrum sensing. The Riemannian manifold neural networks not only fully leverage the advantages of Riemannian manifolds in non-Euclidean data structures, but also combine the powerful expressive capabilities of neural networks, thus demonstrating significantly superior spectrum sensing performance in various complex environments. A series of detailed simulation experiments validate the superior performance of this method in diverse environments, showcasing its potential application value in actual wireless communication systems.更多还原