Multi-task learning approach for modulation and wireless signal classification for 5G and beyond: Edge deployment via model compression

Future communication networks must address the scarce spectrum to accommodate extensive growth of heterogeneous wireless devices. Efforts are underway to address spectrum coexistence, enhance spectrum awareness, and bolster authentication schemes. Wireless signal recognition is becoming increasingly more significant for spectrum monitoring, spectrum management, secure communications, among others. Consequently, comprehensive spectrum awareness on the edge has the potential to serve as a key enabler for the emerging beyond 5G (fifth generation) networks. State-of-the-art studies in this domain have (i) only focused on a single task – modulation or signal (protocol) classification – which in many cases is insufficient information for a system to act on, (ii) consider either radar or communication waveforms (homogeneous waveform category), and (iii) does not address edge deployment during neural network design phase. In this work, for the first time in the wireless communication domain, we exploit the potential of deep neural networks based multi-task learning (MTL) framework to simultaneously learn modulation and signal classification tasks while considering heterogeneous wireless signals such as radar and communication waveforms in the electromagnetic spectrum. The proposed MTL architecture benefits from the mutual relation between the two tasks in improving the classification accuracy as well as the learning efficiency with a lightweight neural network model. We additionally include experimental evaluations of the model with over-the-air collected samples and demonstrate first-hand insight on model compression along with deep learning pipeline for deployment on resource-constrained edge devices. We demonstrate significant computational, memory, and accuracy improvement of the proposed model over two reference architectures. In addition to modeling a lightweight MTL model suitable for resource-constrained embedded radio platforms, we provide a comprehensive heterogeneous wireless signals dataset for public use.     

深海远程正交频分复用水声通信簇约束的分布式 压缩感知信道估计*

在深海远程正交频分复用(Orthogonal Frequency Division Multiplexing,OFDM)水声通信中,信道时延较长,导致信道频率选择性衰落严重,传统的压缩感知信道估计算法性能大幅下降。为此,本文利用深海远程信道在一定时间内具有相关性并呈较稳定簇状分布的特点,在分布式压缩感知信道估计中引入簇区域信息,进行簇约束的多数据块联合稀疏信道估计,提出一种簇约束的分布式压缩感知信道估计方法,并在深海开展了定点远程水声通信实验进行验证,实验结果表明,与传统的分布式压缩感知信道估计算法相比,本方法的能够降低50%左右的误码率。     

Spectrum sensing in cognitive radio networks and metacognition for dynamic spectrum sharing between radar and communication system: A review

The massive growth in mobile users and wireless technologies has resulted in increased data traffic and created demand for additional radio spectrum. This growing demand for radio spectrum has resulted in spectrum congestion and mandated the need for coexistence between radar and interfering communication emitters. To address the aforementioned issues, it is critical to review existing policies and evaluate new technologies that can utilize spectrum in an efficient and intelligent manner. Cognitive radio and cognitive radar are two promising technologies that exploit spectrum using dynamic spectrum access techniques. Additionally, introducing the bio-inspired concept ‘metacognition’ in a cognitive process has shown to increase the effectiveness and robustness of the cognitive radio and cognitive radar system. Metacognition is a high-order thinking agent that monitors and regulates the cognition process through a feedback and control process called the perception–action cycle. Extensive research has been done in the field of spectrum sensing in cognitive radio and spectral coexistence between radar and communication systems. This paper provides a detailed classification of spectrum sensing schemes and explains how dynamic spectrum access strategies share the spectrum between radar and communication systems. In addition to this, the fundamentals of cognitive radio, its architecture, spectrum management framework, and metacognition concept in radar are discussed. Furthermore, this paper presents various research issues, challenges, and future research directions associated with spectrum sensing in cognitive radar and dynamic spectrum access strategies in cognitive radar.     

认知无线电网络中利用FEC和DE的多节点频谱感知算法

针对认知无线电网络(CRN)中空闲频谱感知困难的问题,本文提出了基于前向纠错和差分进化算法的多节点频谱感知算法。首先,利用基于差分进化算法的协同检测完成信号感知;然后,研究了信道噪声对频谱感知性能的影响;最后,分析了前向纠错技术在信道存在噪声时对频谱感知性能的影响。仿真实验将纠错和无纠错控制信道的不同信噪比作为依据,采用三种不同的检测方法评估了本文算法。仿真实验结果表明,在存在噪声的认知无线电网络中,本文算法提高了系统的性能和检测概率,且协同感知算法的性能随着节点数目的增加而提高,该算法适合应用于实时性要求较高的应用程序。     

基于DCGAN的拉曼光谱样本扩充及应用研究

拉曼光谱检测方法依赖于化学计量学算法,深度学习是当下最炙手可热的方向,可应用于拉曼光谱进行建模。但是深度学习需要大样本进行训练,而拉曼光谱采集受制于器材和人力成本,获取大批量的样本需要更大成本,且易受荧光等因素干扰,这些问题都制约了将深度学习应用于拉曼光谱。针对以上问题,通过引入深度卷积生成对抗网络（DCGAN）提取拉曼光谱内部特征,对抗生成新的拉曼光谱,从而达到扩充数据集目的。同时和另一个扩充数据集的方法--偏移法进行对比,证明DCGAN的可靠性。设计生成光谱选取标准,选取高相似性的光谱填充数据集,为深度学习在拉曼光谱中的应用奠定基础。为了验证生成的光谱比原始光谱有更好的适用性,设计四组实验：（1）使用原始拉曼光谱输入到SVM进行分类,得到51.92%的分类准确率;（2）使用原始拉曼光谱输入到CNN进行分类,得到75.00%的分类准确率;（3）采用偏移法生成光谱,输入到CNN里进行分类,得到91.85%的分类准确率;（4）使用DCGAN生成光谱,输入到CNN里进行分类,得到98.52%分类准确率。实验结果表明,DCGAN能在只有少量拉曼光谱的情况下,通过对抗学习得到较好的生成光谱,且生成的光谱相比原光谱更加清晰,减少了可能的干扰因素,具有光谱预处理效果。通过DCGAN对抗生成大量高质量的数据填充到原有拉曼光谱数据集,扩充数据集的样本量,使得深度学习模型能够得到更好的训练,从而提高模型的准确率。该研究为深度学习方法应用于拉曼光谱分析技术提出了一个可行的方案。     

Inverse Design of Dual-Band Optically Transparent Metasurface Absorbers with Neural-Adjoint Method

Massive applications of microwave technology have led to a large amount of electromagnetic pollution, which seriously interferes with the normal operation of communication systems in people's daily life. Optically transparent metasurface absorber is emerging as a promising candidate for solving this problem in some specific scenarios. When designing multiband metasurface absorbers, the coupling within the cell structure generally prevents the independent modulation of the absorption effect in each band. The rapid development of deep learning provides a new way for designing high-performance metasurface. Here, a forward network model is built to predict the absorption spectrum of the metasurface absorber. In the network model, 1D (one-dimensional) inverse convolution is used as the upsampling layer, which enables the network model to have a good prediction on small data sets while avoiding overfitting on large data sets. Based on the network model, a dual-band optically transparent metasurface absorber is also designed by employing the neural-adjoint (NA) method. The results taking advantage of deep learning method for high-efficiency design of metasurface absorbers without considering the coupling effect, have shown great capability to optimize the absorption efficiency of different frequency bands independently, which may provide an alternative way to design other multiband microwave devices.     

Deep learning-driven MIMO: Data encoding and processing mechanism

Large-scale Multiple-Input Multiple Output (MIMO) is the key technology of 5G communication. However, dealing with physical channels is a complex process. Machine learning techniques have not been utilized commercially because of the limited learning capabilities of traditional machine learning algorithms. We design a deep learning hybrid precoding scheme based on the attention mechanism. The method mainly includes channel modeling and deep learning encoding two modules. The channel modeling module mainly describes the problem formally, which is convenient for the subsequent method design and processing. The model design module introduces the design framework, details, and main training process of the model. We utilize the attention layer to extract the eigenvalues of the interference between multiple users through the output attention distribution matrix. Then, according to the characteristics of inter-user interference, the loss minimization function is used to study the optimal precoder to achieve the maximum reachable rate of the system. Under the same condition, we compare our proposed method with the traditional unsupervised learning-based hybrid precoding algorithm, the TTD-based (True-Time-Delay, TTD) phase correction hybrid precoding algorithm, and the deep learning-based method. Additionally, we verify the role of attention mechanism in the model. Extensive simulation results demonstrate the effectiveness of the proposed method. The results of this research prove that deep learning technology can play a driving role in the encoding and processing of MIMO with its unique feature extraction and modeling capabilities. In addition, this research also provides a good reference for the application of deep learning in MIMO data processing problems.     

基于深度学习的自适应光学波前传感技术研究综述

波前传感是自适应光学系统的重要组成部分,在地基大口径望远镜、激光大气传输、无线光通信、激光驱动核聚变等领域发挥了关键作用,同时也常应用于自由曲面的光学测量中。与此同时,深度学习作为一种较为通用的前沿技术,成功在计算机视觉、自然语言处理等众多领域取得了革命性进展。使用深度学习的方法改进自适应光学系统中的波前传感器,以期实现更精准的波前探测,以及适应更复杂的应用场景是自适应光学的发展趋势,也是深度学习应用领域的一个新课题。介绍了深度学习在自适应光学波前传感中的应用现状,主要分析了在相位反演波前传感器和哈特曼波前传感器中的研究特点,并在最后进行了总结和展望。     

分组全连接的近红外光谱定量分析网络

全连接网络作为深度学习中的一种典型结构,几乎在所有神经网络模型中均有出现。在近红外光谱定量分析中,光谱数据样本数量较少,但每个样本的维度高。导致了两个问题：将光谱直接输入网络,网络的参数量会十分庞大,训练模型需要更多的样本,否则模型容易进入过拟合状态;在输入网络前对光谱进行降维,虽解决了网络参数量过大的问题,但会丢失一部分信息,无法充分发挥网络的学习能力。针对近红外光谱的特性,提出了一种分组全连接的近红外光谱定量分析网络GFCN。该网络在传统的两层全连接网络的基础上,用若干个小的全连接层替代第一个全连接层,克服了直接输入光谱导致网络参数量过大的缺点。采用Tecator和IDRC2018数据集对该方法进行测试,同时与全连接网络FCN和偏最小二乘PLS两种方法进行对比。结果显示：在两个数据集上,GFCN预测效果均优于FCN和PLS。在只有少量样本参与建模的情况下,GFCN依然能够保持较高的预测效果。表明,GFCN可以用于近红外光谱的定量分析,并且适应样本较少的场景,具有重要的研究价值和广泛的应用场景。     

Distributed dynamic spectrum access through multi-agent deep recurrent Q-learning in cognitive radio network

This paper addresses the problem of distributed dynamic spectrum access in a cognitive radio (CR) environment utilizing deep recurrent reinforcement learning. Specifically, the network consists of multiple primary users (PU) transmitting intermittently in their respective channels, while the secondary users (SU) attempt to access the channels when PUs are not transmitting. The problem is challenging considering the decentralized nature of CR network where each SU attempts to access a vacant channel, without coordination with other SUs, which result in collision and throughput loss. To address this issue, a multi-agent environment is considered where each of the SUs perform independent reinforcement learning to learn the appropriate policy to transmit opportunistically so as to minimize collisions with other users. In this article, we propose two long short-term memory (LSTM) based deep recurrent Q-network (DRQN) architectures for exploiting the temporal correlation in the transmissions by various nodes in the network. Furthermore, we investigate the effect of the architecture on success rate with varying number of users in the network and partial channel observations. Simulation results are compared with other existing reinforcement learning based techniques to establish the superiority of the proposed method.     

基于贝叶斯网络的认知引擎设计与重配置

以往的通信行为指导系统未来通信, 以满足用户需求并适应环境变化, 是认知无线电系统的核心所在, 为此提出了一种基于贝叶斯网络的认知引擎, 用于解决在复杂多变的电磁环境与用户需求条件下, 认知无线电系统参数自适应调整的问题. 通过对系统过去通信行为样本数据, 进行结构学习和参数学习建立认知引擎, 将系统当前环境状态和用户需求信息经预处理作为推理的证据, 应用引擎决策出系统此时最佳的工作参数, 完成系统参数重构. 本文利用OPNET工具建立一个移动无线网络完成仿真实验, 仿真结果表明该认知引擎能有效地使移动无线网络适应环境变化, 改善端到端通信性能, 进一步验证了建模方法的可行性.     

BeiDou Short-Message Satellite Resource Allocation Algorithm Based on Deep Reinforcement Learning

The comprehensively completed BDS-3 short-message communication system, known as the short-message satellite communication system (SMSCS), will be widely used in traditional blind communication areas in the future. However, short-message processing resources for short-message satellites are relatively scarce. To improve the resource utilization of satellite systems and ensure the service quality of the short-message terminal is adequate, it is necessary to allocate and schedule short-message satellite processing resources in a multi-satellite coverage area. In order to solve the above problems, a short-message satellite resource allocation algorithm based on deep reinforcement learning (DRL-SRA) is proposed. First of all, using the characteristics of the SMSCS, a multi-objective joint optimization satellite resource allocation model is established to reduce short-message terminal path transmission loss, and achieve satellite load balancing and an adequate quality of service. Then, the number of input data dimensions is reduced using the region division strategy and a feature extraction network. The continuous spatial state is parameterized with a deep reinforcement learning algorithm based on the deep deterministic policy gradient (DDPG) framework. The simulation results show that the proposed algorithm can reduce the transmission loss of the short-message terminal path, improve the quality of service, and increase the resource utilization efficiency of the short-message satellite system while ensuring an appropriate satellite load balance.     

Using physics-informed enhanced super-resolution generative adversarial networks for subfilter modeling in turbulent reactive flows

Turbulence is still one of the main challenges in accurate prediction of reactive flows. Therefore, the development of new turbulence closures that can be applied to combustion problems is essential. Over the last few years, data-driven modeling has become popular in many fields as large, often extensively labeled datasets are now available and training of large neural networks has become possible on graphics processing units (GPUs) that speed up the learning process tremendously. However, the successful application of deep neural networks in fluid dynamics, such as in subfilter modeling in the context of large-eddy simulations (LESs), is still challenging. Reasons for this are the large number of degrees of freedom in natural flows, high requirements of accuracy and error robustness, and open questions, for example, regarding the generalization capability of trained neural networks in such high-dimensional, physics-constrained scenarios. This work presents a novel subfilter modeling approach based on a generative adversarial network (GAN), which is trained with unsupervised deep learning (DL) using adversarial and physics-informed losses. A two-step training method is employed to improve the generalization capability, especially extrapolation, of the network. The novel approach gives good results in a priori and a posteriori tests with decaying turbulence including turbulent mixing, and the importance of the physics-informed continuity loss term is demonstrated. The applicability of the network in complex combustion scenarios is furthermore discussed by employing it in reactive and inert LESs of the Spray A case defined by the Engine Combustion Network (ECN).     

基于改进混合蛙跳算法的认知无线电协作频谱感知

提出了一种改进的混合蛙跳算法（shuffled frog leaping algorithm,SFLA）,并提出了基于改进SFLA的认知无线电协作频谱感知方法,通过仿真对改进SFLA算法性能与传统SFLA算法性能进行了比较,并对本文提出的基于改进SFLA的协作感知方法与已有的基于修正偏差因子（modified deflection coefficient,MDC）的协作感知方法性能进行了比较.结果表明改进SFLA算法性能优于传统SFLA;基于改进SFLA的协作感知方法比MDC方法能获得更大的检测概率,验证 关键词： 认知无线电 频谱感知 混合蛙跳算法     

On spectrum sensing in cognitive radio CDMA networks with beamforming

In this paper, the performance of cognitive radio (CR) code division multiple access (CDMA) networks is analyzed in the presence of receive beamforming at the base stations (BSs). More precisely, we analyze, through simulations, the performance achievable by a CR user, with and without spectrum sensing, in a three-cell scenario. Uplink communications are considered. Three different schemes for spectrum sensing with beamforming are presented, together with a scheme without spectrum sensing. CR users belong to a cognitive radio network (CRN) which is coexisting with a primary radio network (PRN). Both the CRN and the PRN are CDMA based. The CRN is assumed to utilize beamforming for its CR users. Soft hand-off (HO) and power control are considered in both the CRN and the PRN. The impact of beamforming on the system performance is analyzed, considering various metrics. In particular, we evaluate the performance of the proposed systems in terms of outage probability, blocking probability, and average data rate of CR users. The results obtained clearly indicate that significant performance improvements can be obtained by CR users with the help of beamforming. The impact of several system parameters on the performance of the three considered spectrum sensing schemes with beamforming is analyzed. Our results, in terms of probability of outage, show that the relative improvement brought by the use of beamforming is higher in the absence of spectrum sensing (reduction of 80%) than in the presence of spectrum sensing (reduction of 42%).     

Spectrum mobility in cognitive radio network using spectrum prediction and monitoring techniques

The spectrum mobility during data transmission is an integral part of the cognitive radio network (CRN) which is conventionally two types for instance reactive and proactive. In the reactive approach, the cognitive user (CU) switches its communication after the emergence of the primary user (PU), where the detection of emergence of PU relies either on spectrum sensing and/or monitoring. Due to certain limitations of the reactive approach such as: (1) loss at least one packet on the emergence of PU and (2) resource (bandwidth) wastage if the periodic sensing is used for mobility, the researchers have introduced the concept of proactive spectrum mobility. In this approach, the emergence of PU is predicted on the bases of pre-available spectrum information, and switching is performed before true emergence of the PU, in order to avoid even the single packet loss. However, the imperfect spectrum prediction is a major milestone for the proactive spectrum mobility. Recently, due to introduction of the spectrum monitoring simultaneous to the data transmission, the reactive approach has come into lime-light again, however, it suffers from the ‘single packet loss’ and ‘imperfect spectrum monitoring’ issues. Therefore in this paper, we have exploited the spectrum monitoring and prediction techniques, simultaneously for the spectrum mobility, in order to enhance the performance of cognitive radio network (CRN). In the proposed strategy, the decision results of the spectrum prediction and monitoring techniques are fused using AND and OR fusion rules, for the detection of emergence of PU during the data transmission. Further, the closed-form expressions of the resource wastage, achieved throughput, interference power at PU and data-loss for the proposed approaches as well as for the prediction and monitoring approaches are derived. Moreover, the simulation results for the proposed approaches are presented and validation is performed by comparing the results with prediction and monitoring approach. In a special case, when the prediction error is zero, the graphs of all metric values overlies the spectrum monitoring approach, which further validates the proposed approach.     

3D卷积自动编码网络的高光谱异常检测

高光谱图像包含丰富的地物光谱信息,在遥感图像领域有着巨大的发展前景.高光谱图像异常检测无需任何先验光谱信息,便可检测出图像中的异常目标.因此,在国防军事和民用领域都有广泛的应用,是现阶段高光谱图像处理领域的研究热点.然而,高光谱图像存在数据复杂、冗余性强、未标记以及样本数量少等特点,这给高光谱图像异常检测带来了很大的挑...     

一维空洞卷积神经网络的矿物光谱分类

矿物光谱综合反映了岩矿的物理化学特性、组分和内部结构特征,已被应用于岩矿识别研究.传统的矿物光谱分类方法需要先对矿物光谱进行预处理,再采用不同方法分析光谱特征,从而实现分类目的.但同时也会造成部分光谱信息丢失,导致最终分类精度不高且操作过程繁琐、效率低下,难以应对日益增长的大数据处理需求.因此,建立一个准确、高效的矿物...     

标志位频谱感知方法研究

文章提出了一种新的标志位频谱感知方法,主要由数据采集和频谱感知两部分组成.前者主要基于标准的压缩感知技术,研究了一种标志位数据采集方法,仅保留测量数据的标志位信息,从而减少了测量数据的存储量.后者基于一致恢复原理和共轭梯度算法来构造频谱感知算法.仿真结果表明,标志位频谱感知方法可以在降低数据采集量及存储量的同时完美获取原始信号的频谱信息.该方法可以应用于无线通信、电子对抗、智能吸波结构以及感知无线电的前端频谱测量设计阶段中. 关键词： 频谱感知 标志位压缩感知 智能吸波结构