Deep learning-based classification algorithms have been used for automatic modulation recognition (AMR). However, most methods only focus on end-to-end mapping and neglect the classic key features. In this paper, signals are enforced with key classification features to propose a novel deep learning model for AMR by learning the shared latent space of the aligned signals and key features (LLAF); this is done to increase the generalizability of the model and to ensure the physical plausibility of the results. To obtain adequate signal representations, an encoder-decoder architecture is proposed to learn the shared latent space, and the architecture is trained to approximate prior label distributions for precise signal classification. Simulation results verify the high recognition accuracy of the proposed LLAF model under different signal-to-noise ratios (SNRs).
Identifying cohesive subgraphs is an important topic in graph theory and complex network analysis. The quasi-clique, as a generalization of clique, can be used to identify functional and structural properties of various networks. In this paper, we study the maximum weighted quasi-clique problem, and propose a local search algorithm for solving the problem. In the algorithm, an iterated local search method is used as the search framework. To find the quasi-clique with the maximum total weights, hybrid vertex selection strategies are proposed and incorporated into our algorithm. The hybrid strategies utilize a probability-based mechanism for choosing sub-strategies in each round of the local search. We conduct experiments on synthetic networks and real-world networks to show the effectiveness of our algorithm. The results indicate that hybrid strategies perform better than existing methods, and thus our algorithm has a good ability to tackle various networks.
In this letter, the issue of mitigating strong co-channel interference (CCI) in communication systems is addressed. Unlike conventional model-based methods, a novel data-driven scheme is proposed. A recurrent neural network (RNN) is trained to directly demodulate the desired signal under strong CCI. Instead of inputting the original received signal, in-phase and quadrature interference-robust features (IRF) are extracted through preprocess. The RNN is then trained offline to implement sequence labelling, with the IRF sequences and known code sequences of the desired signal as inputs and ground-truth labels. Meanwhile, a guard zone is introduced when loading the IRF sequences to enable better contextual information exploitation by the RNN demodulator. Online tests validated the low bit error rate (BER) of the RNN demodulator, under strong CCI. Moreover, the proposed scheme outperformed existing model-based and data-driven interference mitigation schemes in terms of the BER, especially in low signal-to-interference ratio region. Inspiringly, the proposed data-driven scheme generalized well to varied unseen test conditions.
The improvement of the performance of a distributed Bragg reflector laser bar emitting near 905 nm through the use of multiple epitaxially stacked active regions and tunnel junctions is reported. The bar consisting of 48 emitters (each having an aperture of 50 μm) emits an optical power of 2.2 kW in 8 ns long pulses at an injection current of 1.1 kA. This corresponds to an almost threefold increase of the pulse power compared to a bar with lasers having only a single active region. Due to the integrated surface Bragg grating, the bar exhibits a narrow spectral bandwidth of about 0.3 nm and a thermal tuning of only 68 pm/K.
In this letter, a novel sext-band bandpass filter (BPF) using two split-type multiple-mode resonators is proposed. The filter is composed of two sets of resonators which are respectively designed to realize Filter 1 with two passbands and Filter 2 with quad passbands. Then the two BPFs are combined by parallel coupling feed lines for sext-band responses, and each of the six centre frequencies in the proposed sext-band BPF is able to be controlled due to the design freedom. To validate the design and analysis, a prototype filter has been fabricated with six passbands centered at 1.88/2.59/3.48/5.26/5.82/6.75 GHz. The measured result of the fabricated filter agrees well with the simulation, which shows that the proposed structure is a good candidate for sext-band BPF designs and validates the proposed design flow well.
An ultra-low profile and high-performance UHF RFID reader antenna is proposed in this letter, which can be switched between far-field (FF) and near-field (NF) operative mode. This antenna is composed of four dipoles and a reconfigurable feed network. The four dipoles form a square, and the feed network is located in the center. The feed network is a four-way power divider, and the phase of the output signal can be controlled by switching diodes. As a balanced structure, double-sided parallel-strip lines (DSPSLs) are used to feed four dipoles. Compared to the balun structure, the application of DSPSLs can reduce the area of the feed network, thereby reducing the influence of the feed network on the central magnetic field. Experimental results show that the proposed antenna can provide a NF reading area of 180×180 mm2, and the identification rate can reach 100% within 30-60 mm height. The FF gain of the antenna is 4.7 dBic, and the overlapping range of 3 dB AR bandwidth and -10 dB impedance bandwidth is 800-960 MHz. The good FF and NF performance of this antenna is conducive to its application in RFID systems.
Genomics data analysis requires efficient tools to address the vast amount of data generated by current next-generation sequencing technologies. K-mer counting works face difficulties in balancing high memory overhead with statistical precision. We designed a high-frequency k-mer statistical computation based on the Space Saving algorithm and a novel hash table structure, which reduces the memory overhead by 46\% while ensuring high computational efficiency.
With the increasing penetration of renewable energy, virtual power plants reduce the impact on the power grid by integrating massive distributed resources for unified management. However, the optimal scheduling of a large number of distributed resources in virtual power plants has become a new problem in recent years. Therefore, aiming at the real-time optimal scheduling problem in the optimal scheduling of virtual power plant, this letter regards the virtual power plant as a multi-agent system, and proposes a novel real-time active power dispatch scheme of virtual power plant based on distributed model predictive control, so that each agent can not only calculate its own optimization function relatively independently, but also fully refer to the neighbor information. Simulation results show the feasibility and effectiveness of the proposed method.
The accurate prediction of radio wave propagation is extremely important for wireless network planning and optimization. However, inexact matching between the traditional empirical model and actual propagation environments, as well as the insufficiency of the sample data required for training a deep learning model, lead to unsatisfactory prediction results. Our paper proposes a field strength prediction model based on a deep neural network that is aimed at a tiny dataset composed of the geographic information and corresponding satellite images of a target area. This model connects two pretrained networks to minimize the parameters to be learned. Simultaneously, we construct a convolutional neural network (CNN) model for comparison based on a previous advanced study in this field. Experimental results show that the proposed model can obtain the same accuracy as that of previously developed CNN models while requiring less data.
This paper proposes a joint optimization method for the imaging algorithm and sampling scheme of sparse spotlight syhthetic aperture radar (SAR) imaging based on deep convolutional neural networks. Traditional compressed sensing (CS) based sparse SAR imaging has been widely studied. Deep learning and sparse unfolding networks have been introduced into sparse SAR imaging, but most current works focus only on the imaging stage and simply adopt the conventional uniform or random down-sampling scheme. Considering that the imaging quality also depends on the sampling pattern besides the imaging algorithm, this paper introduces a learning-based strategy to jointly optimize the sampling scheme and the imaging network parameters of the reconstruction module. In a deep learning-based image reconstruction scheme, joint and continuous optimization of the sampling patterns and convolutional neural network parameters is achieved to improve the image quality. Simulation results based on real SAR image dataset illustrate the effectiveness and superiority of the proposed framework.
This paper shows the comparison between several well-known classification algorithms in Machine Learning with the purpose to find the most suitable algorithm to predict the dwelling time i.e., how long a certain tourist should stay in a particular tourist spot. This dwelling time prediction can be adopted for tour and travel agents to provide optimal scheduling for their package tour. The algorithm in question is strictly for classification because in this case, the dwelling time does not require a very specific number of minutes, thus the time can be classified and restricted into several time frames. The origin and features of the dataset are described in this paper as well as the comparison methodology to show the procedure of how the comparison was made. Lastly, the performance results will be used to determine which algorithm to use for this specific case and it will be shown in a form of a graph
In this letter, the jamming resource allocation problem of distributed jammers cooperatively jamming netted radar system is investigated. A well-constructed jamming resource allocation model considering jamming beams, jamming power and other influencing factors is established. Random keys are used in this letter to improve the coding mode of genetic algorithm. Simulation results show that in the case of limited jamming resources, the model and algorithm proposed can achieve effective jamming allocation schemes facing a netted radar with any number of radar nodes.
Rydberg-atom electrometers have the remarkable advantages of self-calibration and high sensitivity. Based on the classical electromagnetic theory, a localized electric field enhancement structure of a hybrid rectangular resonator is proposed to improve the sensitivity of quantum microwave measurement. It should be noted that the prototype of the hybrid rectangular resonator is fabricated and measured at 9.925 GHz. The results of full-wave simulations show that the uniform and high electric field enhancement in the TE101 fundamental mode is realized. The transient process of resonance is simultaneously simulated, and the time to settle steady state is given as about 104 ns. As indicated through experimental results that the structure can reach 24 dB (enhancement factor of 15.8). As a result, the method proposed in this study, based on atomic measurement capabilities, enables us to improve the measurement sensitivity further and promotes the practical development of quantum microwave measurement technology.
Affective video content analysis is an active topic in the field of affective computing. In general, affective video content can be depicted by feature vectors of multiple modalities, so it is important to effectively fuse information. In this work, a novel framework is designed to fuse information from multiple stages in a unified manner. In particular, a unified fusion layer is devised to combine output tensors from multiple stages of the proposed neural network. With the unified fusion layer, a bidirectional residual recurrent fusion block is devised to model the information of each modality. Moreover, the proposed method achieves state-of-the-art performances on two challenging datasets, i.e., the accuracy value on the VideoEmotion dataset is 55.8%, and the MSE values on the two domains of EIMT16 are 0.464 and 0.176 respectively. The code of UMFN is available at: https://github.com/yunyi9/UMFN.
Event-based cameras are sensitive to brightness changes and can capture rich temporal information with very high temporal resolution, which has great potential for motion segmentation of moving objects. Under static background, events are only triggered by motion of objects, thereby moving objects can be easily segmented. However, in many real-world applications, events are also be triggered by the motion of camera or background and submerge the ones corresponding to moving objects. In this letter, we propose an event-based motion segmentation method to segment moving small objects in events obtained from the wild. First, motion estimation is performed to align the events triggered by the background. Then, candidate events corresponding to moving objects or moving backgrounds are detected. Finally, motion information is adopted to segment the events of moving small objects from the ones triggered by the background. In addition, we develop the first dataset for event-based motion segmentation of small objects, namely EMSS. Experimental results demonstrate the effectiveness of our method and show that our method can achieve robust motion segmentation of small moving objects in the wild.
In this letter, a joint weighted power detector (JWPD) based on maximum a posterior probability (MAP) criteria is proposed for Willie aiming at two-hop covert communication scenario, which is a near optimal detector. Instead of only supervising one single phase, Willie combines the observations of two phases to make joint decision in the proposed scheme. The proposed scheme achieves lower probability of detection error (PDE) than the existing single-phase-detector (SPD) scheme and adding-power-directly-detector (APDD) scheme due to sufficient utilization of the two-phases observations. Numerical results demonstrate the benefit of our proposed scheme.
To improve the diversity and performance of the Mayfly Algorithm (MA), this letter adopts the mutation strategies in the process of MA. The opposition-based learning (OBL) and Cauchy mutation strategies are used to mutate the global optimal solution, and the artificial mutation operator is used in the offspring population. The hybrid mutation strategies are used in a cascaded structure. The performance of the proposed algorithms is demonstrated in simulations comparatively.
Pavement distress classification is a vital step for automatic pavement inspection and maintenance. Recently, patch-based approaches have achieved promising performances and thus extensive attention in this field. However, these methods simply assume that all patches contribute equally to the distress classification, leading to weakly discriminating abilities of models. Moreover, their tedious processes also leads to a low efficiency in inference. In this letter, we present a novel patch-based pavement distress classification approach named Deep Patch Soft Selective Learning (DPS$^2$L), which addresses these issues. Similar to other patch-based approaches, DPS$^2$L partitions the pavement images into patches and aggregates the patch features to accomplish the task. To address the first issue, we introduce a succinct Soft Patch Feature Selection Network (SPFSN) to assess the importance of each patch to the distress classification with a score based on its feature. These scores will be considered as patch-wise weights for feature aggregation. In such a manner, the most discriminative patches are selected in a soft way, and thereby benefit the final classification. To address the inference efficiency issue, knowledge distillation is leveraged to transfer the classification knowledge from DPS$^2$L to the image-based approaches, such as EfficientNet-B3. This distilled model enables incorporating both the advantages of patch-based approaches in classification performance and the advantages of image-based approaches in inference efficiency. Extensive experiments on a large-scale pavement image dataset named CQU-BPDD demonstrates the superiority of our methods over baselines regardless of performance or efficiency.