Automatic modulation classification (AMC), which aims to blindly identify the modulation type of an incoming signal at the receiver in wireless communication systems, is a fundamental signal processing technique in the physical layer to improve the spectrum utilization efficiency. Motivated by deep learning (DL) high-impact success in many informatics domains, including radio signal processing for communications, numerous recent AMC methods exploiting deep networks have been proposed to overcome the existing drawbacks of traditional approaches. DL is capable of learning the underlying characteristics of radio signals effectively for modulation pattern recognition, which in turn improves the modulation classification performance under the presence of channel impairments. In this work, we first provide the fundamental concepts of various architectures, such as neural networks, recurrent neural networks, long short-term memory, and convolutional neural networks as the necessary background. We then convey a comprehensive study of DL for AMC in wireless communications, where technical analysis is deliberated in the perspective of state-of-the-art deep architectures. Remarkably, several sophisticated structures and advanced designs of convolutional neural networks are investigated for different data types of sequential radio signals, spectrum images, and constellation images to deal with various channel impairments. Finally, we discuss some primary research challenges and potential future directions in the area of DL for modulation classification.

Automotive radars, with a widespread emergence in the last decade, have faced various jamming attacks. Utilizing low probability of intercept (LPI) radar waveforms, as one of the essential solutions, demands an accurate waveform recognizer at the intercept receiver. Numerous conventional approaches have been studied for LPI radar waveform recognition, but their performance is inadequate under channel condition deterioration. In this letter, by exploiting deep learning (DL) to capture intrinsic radio characteristics, we propose a convolutional neural network (CNN), namely LPI-Net, for automatic radar waveform recognition. In particular, radar signals are first analyzed by a time-frequency analysis using the Choi-Williams distribution. Subsequently, LPI-Net, primarily consisting of three sophisticated modules, is built to learn the representational features of time-frequency images, in which each module is constructed with a preceding maps collection to gain feature diversity and a skip-connection to maintain informative identity. Simulation results show that LPI-Net achieves the 13-waveform recognition accuracy of over 98% at 0 dB SNR and further performs better than other deep models.

This paper studies the performance of intelligent reflecting surface (IRS)-aided short-packet non-orthogonal multiple access (NOMA) systems under perfect and imperfect successive interference cancellation. Closed-form expressions for the average block error rate (BLER) with random and optimal phase shifts are derived. To gain some useful insights, an asymptotic analysis is carried out, based on which the diversity gain is achieved. Numerical results show that, at high signal-to-noise ratio, users with random phase shift achieve significant performance improvements in the performance gap between direct and non-direct links from a source when compared with that of the optimal phase shift. In addition, the proposed system outperforms its orthogonal multiple access counterpart in terms of both BLER and throughput. Furthermore, optimal channel coding rates of users can maximize the system throughput performance. Monte Carlo simulations are presented to corroborate the theoretical analysis.

In recent years, the increasing popularity of unmanned aerial vehicles (UAVs) has arisen from the emergence of cutting-edge technologies deployed in small and low-cost devices. With the great capability of friendly uses and wide applications for multiple purposes, amateur drones can be piloted to effortlessly access any geographical area. This poses some difficulties in monitoring and managing drones that may invade private or limited-access areas. In this paper, we propose a radio-frequency (RF)-based surveillance solution to effectively detect and classify drones, and recognize operations by leveraging a high-performance convolutional neural network. The proposed network, namely RF-UAVNet, is specified with grouped one-dimensional convolution to significantly reduce the network size and computational cost. Besides, a novel structure of multi-level skip-connection, for the preservation of gradient flow, incorporating multi-level pooling, for the collection of informative deep features, is proposed to achieve high accuracy via learning efficiency improvement. In the experiments, RF-UAVNet yields the accuracy of 99.85% for drone detection, 98.53% for drone classification, and 95.33% for operation mode recognition, numbers which outperform the current state-of-the-art deep learning-based methods on DroneRF, a publicly available dataset for RF-based drone surveillance systems.

In this article, we propose a cooperative nonorthogonal multiple access (NOMA)-enabled simultaneous wireless information and power transfer (SWIPT) Internet of Things (IoT) networks, where one information source harvests energy from a multiantennas power beacon (PB) to serve two IoT users via the help of multiple energy-limited relay nodes. To improve the performance of far IoT user, we propose reactive and proactive relay selection protocols together with time-power energy harvesting mechanism under imperfect successive interference cancellation. Closed-form expressions for the outage probability (OP), throughput, and energy efficiency (EE) of the proposed system are obtained, from which the asymptotic analysis for the throughput is also carried out. To further enhance the system performance, we propose a low-complexity method to optimize the outage and throughput performance subject to power allocation, time, and power splitting parameters. Toward real-time configurations in IoT networks, we design a deep learning framework for the sum-throughput and EE predictions with low computation complexity and high accuracy. The influences of antennas setting at PB, time-switching ratio, power-splitting ratio, power allocation factor, and the number of relays on the system OP, throughput, and EE are evaluated and discussed along with numerical results.

In this paper, we study short-packet communications in wireless-powered cognitive Internet-of-Things (IoT) networks with multiple primary receivers (PRs). The considered system can be applied for small factory automations, where a source and multiple relays harvest energy from a multi-antenna dedicated power beacon (PB) to send short packets to a robot destination for controlling purposes under cognitive radio constraint imposed by PRs. We propose an opportunistic relay selection (ORS) scheme to maximize the end-to-end signal-to-noise ratio in cognitive IoT systems. Closed-form expressions for the average block error rate (BLER) of the proposed system are obtained, based on which the performance floor analysis, goodput, and energy efficiency (EE) are also carried out. Relying on analytical results, we develop a deep learning framework for the BLER prediction with high accuracy and short execution time. Simulation results show the BLER, goodput, and EE improvements of the ORS scheme over conventional relay selection schemes. Moreover, the developed deep learning-based evaluation model achieves the equivalent performance as the ORS scheme in terms of BLER, goodput, and EE, while remarkably reducing the execution time in cognitive IoT systems.