ID 原文 译文
58268 针对不同轨道角动量( OAM) 叠加的涡旋光束探测问题，提出了基于机器学习的模式识别技术，为 OAM 叠加光束的检测提供了一个新思路.基于修正的 von Karman 功率谱模型，利用功率谱反演法生成随机相位屏，应用多步衍射法数值模拟拉盖尔高斯叠加光束在大气湍流信道的传输.研究了不同波长、传输距离和大气湍流信道条件下训练的卷积神经网络( CNN) 分别对各种湍流强度测试集的识别正确率. In order to solve the problem of detection of different orbital angular momentum ( OAM) superimposed vortex beams，a pattern recognition technology based on machine learning ( ML) is proposed，which provides a brand-new method for multi-OAM states detection. In order to study the recognition rateof multi-OAM beams using convolutional neural network ( CNN) models under different wavelength，transmission distance and atmospheric turbulence conditions，the numerical simulation phase screens aregenerated by the power spectral inversion method based on the modified von Karman power spectrummodel.
58269 结果表明:对于较弱的湍流、波长较长的 OAM 光束和较短的传输距离条件，基于 CNN 的 OAM 模式识别正确率较高; Multi-step diffraction method is used to simulate numerically the propagation of OAM beams in theatmospheric turbulence，and the training and testing database are obtained under different atmosphericturbulence. Ｒesults indicate the accuracy of CNN-based OAM pattern recognition increases as wavelengthincreases，transmission distance decreases and turbulent intensity decreases. And the CNN trained withthe database under strong turbulence has high accuracy for all kind of turbulence condition，and usingmixed training database under different turbulence condition can enhance the accuracy.
58270 对于各种湍流条件的测试数据，使用强湍流训练集训练得到的模型与使用弱湍流训练集训练得到的模型相比识别正确率更高;利用混合训练集进行训练有利于提高识别正确率.这些结果对 OAM 光束解复用系统的实现具有一定的参考价值. These resultscontribute to the demultiplexing systems of free space optical-OAM systems.
58271 针对频谱资源中授权频段因信道质量和信号干扰等因素导致的感知频谱利用率不高的问题，提出了基于大规模天线空间信道控制的频谱感知资源分配方式. Aiming at problem of lower utilization the authorized spectrum，due to channel quality andtransmission power interference，a resource allocation technology was proposed based on the large scaleantenna spatial channel controlling.
58272 通过波束成型技术使主瓣波束对准期望用户，而在干扰源方向放置零陷，可有效地抑制对次级用户的干扰，并确保次级用户对感知频谱的使用，从而进一步提高系统的频谱效率. Main lobes beam of multiple antenna beam forming points to the desired user and places the nulls in the interference direction，which can effectively suppress interference tosecondary users and ensure that aware spectrum is used. As a result，the utilization ratio of frequencyspectrum increased further in the whole system.
58273 仿真结果显示，在信噪比为 5 dB 和 10 dB 时，所提方案的谱效率( 系统吞吐量) 优于其他策略. The simulation results showed the throughput of the proposed method is better than the scheme in the literatures when signal to noise ratio is 5 dB and 10 dB.
58274 核四元数主成分分析( KQPCA) 被成功应用于处理非线性四元数信号，然而，核矩阵维数太高使其对角化非常耗时，目前二维形式的 KQPCA( 2DKQPCA) 并没有成功实现. Currently，kernel quaternion principal component analysis ( KQPCA) has been proposed andsuccessfully applied to process linear quaternion signals. However，two dimensional version of KQPCA( 2DKQPCA) has not been successfully implemented due to the quite time-consuming problem for diagonalizing the high dimensional kernel matrix.
58275 对此，采用基于块处理和并行计算的思想，提出基于块的 2DKQPCA( B2DKQPCA) ，实现真正意义上的 2DKQPCA. So，using the block-based idea and the parallel computingidea，the block-wise 2DKQPCA ( B2DKQPCA) is proposed to implement 2DKQPCA really.
58276 基于时间复杂度、应用性能和分块矩阵应为四元数Hermitian 矩阵的综合考虑，B2DKQPCA 重点处理主对角线、反对角线和主对角线旁 3 个方向的小块. After theoverall consideration of computational complexity，application performance and quaternion Hermitianblock，B2DKQPCA mainly processes the blocks of three directions: main-diagonal direction，anti-diagonal direction and side-diagonal direction.
58277 然后，结合B2DKQPCA 与 ＲGB-D 图像四元数表示方法，将 B2DKQPCA 应用于 ＲGB-D 目标识别领域.在 2 个公开库上的实验结果表明，提出的基于列向 B2DKQPCA 的 ＲGB-D 识别算法优于现有基于主成分分析算法和基于卷积神经网络的一些算法. Then，B2DKQPCA is applied into ＲGB-D object recognition bycombining B2DKQPCA and quaternion representation of ＲGB-D images. Experimental results on twopublicly available datasets demonstrate that the proposed ＲGB-D object recognition algorithm based on thecolumn direction B2DKQPCA outperforms some existing algorithms using principal component analysis and some existing algorithms using convolutional neural network.