基于深度学习的认知物联网频谱感知算法研究  

Research on Spectrum Sensing Algorithm of Cognitive Internet of Things Based on Deep Learning

作　　者：王安义[1] 王文龙 梁艳 WANG Anyi;WANG Wenlong;LIANG Yan(College of Communication and Information Technology,Xi'an University of Science and Technology,Xi'an 710054,China)

机构地区：[1]西安科技大学通信与信息工程学院,陕西西安710054

出　　处：《无线电工程》2024年第3期679-686,共8页Radio Engineering

基　　金：国家自然科学基金(U19B2015)。

摘　　要：针对认知物联网(Internet of Things, IoT)对低信噪比(Signal to Noise Ratio, SNR)的频谱感知性能低下以及传统卷积神经网络(Convolutional Neural Network, CNN)频谱感知方法提取数据特征不充分导致感知性能差等问题,提出了一种改进残差网络——ResNeXt的单节点频谱感知算法,ResNeXt只需要设置少量超参数且高度模块化,将该网络在图像处理上的优势应用在频谱感知问题上,先将接收信号转成二维矩阵并归一灰度化处理,得到灰度图像作为网络的输入。通过训练ResNeXt来提取灰度图像特征,将在线数据输入完成频谱感知。将各个次用户(Secondary User, SU)得到的评分向量矩阵直接用融合中心SoftCombinationNet(SCN)融合获得协作频谱感知结果,有效解决了传统硬融合方法检测性能低、软融合处理复杂等问题。实验结果表明,所提方法在低SNR仍能实现低虚警率、高检测概率,优于传统频谱感知方法。To address the problems of low performance of spectrum sensing in low Signal to Noise Ratio(SNR)and poor sensing performance due to inadequate extraction of data features by traditional Convolutional Neural Network(CNN)spectrum sensing methods for cognitive Internet of Things(IoT).A single-node spectrum sensing algorithm with an improved Residual Network ResNeXt is proposed,which requires only a small number of hyperparameters and is highly modular.The advantages of this network in image processing are applied to the spectrum sensing problem by first converting the received signal into a two-dimensional matrix and normalizing the grayscale processing to obtain a grayscale image as the input to the network.By training ResNeXt to extract grayscale image features,the online data are input to complete the spectrum sensing.Finally,the score vector matrix obtained from each Secondary User(SU)is directly fused with the fusion center SoftCombinationNet(SCN)to obtain the collaborative spectrum sensing results,which effectively solves the problems of low detection performance of traditional hard fusion methods and complex processing of soft fusion.The experimental results show that the proposed method can achieve low false alarm rate and high detection probability even at low SNR,which is better than the traditional spectrum sensing method.

关 键 词：频谱感知 认知物联网 深度学习 协作频谱感知 

分 类 号：TN911[电子电信—通信与信息系统]