Search

Ground-penetrating radar (GPR) has been widely applied to the detection and delineation of buried targets in the subsurface. Compared with conventional single-channel GPR, polarimetric GPR has been proven to possess an improved ability to detect and characterize an elongated object in the subsurface. This letter proves that the scattering signals from a cylinder in two orthogonal polarization channels have a phase difference of about 90° when its diameter-to-wavelength ratio is about 0.05–0.33. Consequently, a polarization-difference imaging method, which shifts the VV component by −90° and subtracts it from the HH component, is proposed for the improved detection and imaging of a subsurface elongated object. Its effectiveness is verified by numerical, laboratory, and field tests on buried rebars and pipes. The signal-to-clutter ratios of the reconstructed GPR images can be improved by up to 4.5 dB by considering the phase difference between the dual-polarization components. [ABSTRACT FROM AUTHOR]

To meet the demand for simultaneously supporting the microwave and millimeter-wave technologies in emerging industrial wireless systems, research on dual-band components with a large frequency ratio has been conducted recently. However, it is challenging to implement independent frequency tunability on bandpass filters with both compactness and high performance. This article presents a dual-band bandpass filter with a large frequency ratio, independently tunable frequencies, and an ultrawide stopband. This is achieved by artfully incorporating a harmonic-controllable varactor-loaded microstrip BPF and a wide-stopband dielectric-loaded cavity BPF with the bandgap-embedded aperture coupling mechanism. Extremely compact size can be achieved without additional filtering elements. Two pairs of transmission zeros can be allocated beside two passbands for good selectivity due to the separate path and cross-coupling in the resonators. For validation, a prototype is designed, fabricated and measured. It can provide two independent tunable frequency ranges (3.22–3.52 GHz and 25–27.06 GHz), exhibiting a large frequency ratio up to 8.3. Mid-band suppression better than 20 dB and ultra-wide stopband up to 32.5 GHz with 20 dB suppression level are achieved. Low insertion loss and good selectivity can also be observed. [ABSTRACT FROM AUTHOR]

This article presents a rough-to-fine evolutionary multiobjective optimization algorithm based on the decomposition for solving problems in which the solutions are initially far from the Pareto-optimal set. Subsequently, a tree is constructed by a modified $k$ -means algorithm on $N$ uniform weight vectors, and each node of the tree contains a weight vector. Each node is associated with a subproblem with the help of its weight vector. Consequently, a subproblem tree can be established. It is easy to find that the descendant subproblems are refinements of their ancestor subproblems. The proposed algorithm approaches the Pareto front (PF) by solving a few subproblems in the first few levels to obtain a rough PF and gradually refining the PF by involving the subproblems level-by-level. This strategy is highly favorable for solving problems in which the solutions are initially far from the Pareto set. Moreover, the proposed algorithm has lower time complexity. Theoretical analysis shows the complexity of dealing with a new candidate solution is $\mathcal {O}(M \log N)$ , where $M$ is the number of objectives. Empirical studies demonstrate the efficacy of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Head pose estimation (HPE) has wide industrial applications, such as online education, human–robot interaction, and automatic manufacturing. In this article, we address two key problems in HPE based on label learning and asymmetric relation cues: 1) how to bridge the gap between the better prediction performance of networks and incorrectly label pose images in the HPE datasets and 2) how to take full advantage of the adjacent poses information around the centered pose image. We reconstruct all the incorrect labels as a two-dimensional Lorentz distribution to tackle the first problem. Instead of directly adopting the angle values as hard labels, we assign part of the probability values (soft labels) to adjacent labels for learning discriminative feature representations. To address the second problem, we reveal the asymmetric relation nature of HPE datasets. The yaw direction and pitch direction are assigned different weights by introducing the half at half-maximum of the Lorentz distribution. Compared with the traditional end-to-end frameworks, the proposed one can leverage the asymmetric relation cues for predicting the head pose angle in the incorrect label scenarios. Extensive experiments on two public datasets and our infrared dataset demonstrate that the proposed ARHPE network significantly outperforms other state-of-the-art approaches. [ABSTRACT FROM AUTHOR]

Although product design has been considered as a collaborative activity, progress in developing methods to facilitate the design process has been hampered by difficulties in building common bases among various stakeholders. Customers may only have general needs of the product in layman's terms instead of the sufficient domain knowledge to identify the product specifications. Thus, there is a great need in design research to translate the expressed needs in natural language to design specifications and bridge this semantic gap. By leveraging the massive amount of online user-generated content, we develop a deep learning-based method to automatically identify product design parameters or product specifications. Specifically, we crawl product review data and the corresponding product metadata from e-commerce websites. A convolutional neural network based solution is provided to map product reviews to product specifications. Experimental results show that the method can be well adapted to the mapping from customer needs in the natural language to product specifications. The method facilitates product design by addressing the semantic gap between general customer needs and detailed product specifications. In addition, the automated mapping greatly improves the efficiency and reduces labor in the design process. [ABSTRACT FROM AUTHOR]

This paper describes a proof-of-concept sensor (1) that can detect and record (without battery) a semiconductor-chip tamper activity (i.e., de-soldering followed by mechanical banging) which a counterfeiter does to scavenge semiconductor chips from a printed circuit board and (2) that can be wirelessly interrogated without need to open semiconductor packages. The sensor is based on a High-overtone-Bulk-Acoustic-Resonator (HBAR) working as a Radio-Frequency Identification (RFID) tag, which can be permanently broken down by the voltage and charge generated by a pyroelectric-energy-converter (PEC). The concept is demonstrated through connecting a 7.5 GHz HBAR (based on a 350 nm-thick ZnO thin film deposited on sapphire) to a $5\times 5\times0.15$ mm3 lithium niobate-based PEC, via a Kapton-copper cantilever switch. When the switch is turned on (due to a mechanical shock emulating a counterfeiter’s mechanical banging), the accumulated charge on the PEC (produced by 250 °C temperature rise emulating a counterfeiter’s de-soldering) generates an electrical pulse with 9.6 V peak voltage and 1.36 nC charge transfer to an 8 $\text{M}\Omega $ load (close to the HBAR’s resistance). The PEC’s voltage and charge are shown to permanently breakdown the ZnO film on the HBAR ($0.1\times 0.1\times0.33$ mm3) having a quality factor (Q) of about 2,500, so that the Q drops to near zero, drastically changing the RF absorption/scatter characteristics. A wireless interrogation is demonstrated by having a pair of $2\times 2\times0.2$ cm3 microstrip patch antennas, one of which is connected to HBAR while the other one is connected to a network analyzer working as a wireless interrogator. [ABSTRACT FROM AUTHOR]

Recommendation accuracy is a fundamental problem in the quality of the recommendation system. In this article, we propose an efficient deep matrix factorization (EDMF) with review feature learning for the industrial recommender system. Two characteristics in user’s review are revealed. First, interactivity between the user and the item, which can also be considered as the former’s scoring behavior on the latter, is exploited in a review. Second, the review is only a partial description of the user’s preferences for the item, which is revealed as the sparsity property. Specifically, in the first characteristic, EDMF extracts the interactive features of onefold review by convolutional neural networks with word-attention mechanism. Subsequently, ${L}_{0}$ norm is leveraged to constrain the review considering that the review information is a sparse feature, which is the second characteristic. Furthermore, the loss function is constructed by maximum a posteriori estimation theory, where the interactivity and sparsity property are converted as two prior probability functions. Finally, the alternative minimization algorithm is introduced to optimize the loss functions. Experimental results on several datasets demonstrate that the proposed methods, which show good industrial conversion application prospects, outperform the state-of-the-art methods in terms of effectiveness and efficiency. [ABSTRACT FROM AUTHOR]

Knowledge graphs are large graph-structured knowledge bases with incomplete or partial information. Numerous studies have focused on knowledge graph embedding to identify the embedded representation of entities and relations, thereby predicting missing relations between entities. Previous embedding models primarily regard (subject entity, relation, and object entity) triplet as translational distance or semantic matching in vector space. However, these models only learn a few expressive features and hard to handle complex relations, i.e., 1-to-N, N-to-1, and N-to-N, in knowledge graphs. To overcome these issues, we introduce a multi-scale dynamic convolutional network (M-DCN) model for knowledge graph embedding. This model features topnotch performance and an ability to generate richer and more expressive feature embeddings than its counterparts. The subject entity and relation embeddings in M-DCN are composed in an alternating pattern in the input layer, which helps extract additional feature interactions and increase the expressiveness. Multi-scale filters are generated in the convolution layer to learn different characteristics among input embeddings. Specifically, the weights of these filters are dynamically related to each relation to model complex relations. The performance of M-DCN on the five benchmark datasets is tested via experiments. Results show that the model can effectively handle complex relations and achieve state-of-the-art link prediction results on most evaluation metrics. [ABSTRACT FROM AUTHOR]