Search

Deep learning methods have been very successful at radio frequency fingerprinting tasks, predicting the identity of transmitting devices with high accuracy. We study radio frequency fingerprinting deployments at resource-constrained edge devices. We use structured pruning to jointly train and sparsify neural networks tailored to edge hardware implementations. We compress convolutional layers by a $27.2\times$ 27. 2 × factor while incurring a negligible prediction accuracy decrease (less than 1 percent). We demonstrate the efficacy of our approach over multiple edge hardware platforms, including a Samsung Gallaxy S10 phone and a Xilinx-ZCU104 FPGA. Our method yields significant inference speedups, $11.5\times$ 11. 5 × on the FPGA and $3\times$ 3 × on the smartphone, as well as high efficiency: the FPGA processing time is $17\times$ 17 × smaller than in a V100 GPU. To the best of our knowledge, we are the first to explore the possibility of compressing networks for radio frequency fingerprinting; as such, our experiments can be seen as a means of characterizing the informational capacity associated with this specific learning task. [ABSTRACT FROM AUTHOR]

Beam selection for millimeter-wave links in a vehicular scenario is a challenging problem, as an exhaustive search among all candidate beam pairs cannot be assuredly completed within short contact times. We solve this problem via a novel expediting beam selection by leveraging multimodal data collected from sensors like LiDAR, camera images, and GPS. We propose individual modality and distributed fusion-based deep learning (F-DL) architectures that can execute locally as well as at a mobile edge computing center (MEC), with a study on associated tradeoffs. We also formulate and solve an optimization problem that considers practical beam-searching, MEC processing and sensor-to-MEC data delivery latency overheads for determining the output dimensions of the above F-DL architectures. Results from extensive evaluations conducted on publicly available synthetic and home-grown real-world datasets reveal 95% and 96% improvement in beam selection speed over classical RF-only beam sweeping, respectively. F-DL also outperforms the state-of-the-art techniques by 20-22% in predicting top-10 best beam pairs. [ABSTRACT FROM AUTHOR]

River width can reveal the water extent and water flow on the earth’s surface. Remote sensing facilitates river width extraction at large scales in an automatic way. This letter proposes a new method named GrabRiver that implements whole-process automation from image preparation to river width calculation. It also, for the first time, develops river graph as the simultaneous river topology for width extraction, combining river planform morphology and network. Three major steps are proposed: 1) mapping and connecting rivers. The Multispectral Water Index (MuWI) is used to produce high-accuracy water maps from imagery, whereby specialized algorithms are used to reduce the impacts from nonriver water (lakes, reservoirs, wetlands) and on-channel objects (bridges, dams, ships) and to enforce river connectivity; 2) Constructing the river graph. A connected river map is skeletonized into the graph while maintaining georeferences as properties of edges and nodes in the river graph, and it is followed by river graph pruning to remove false and redundant river tributaries in the topologic structure; and 3) Measuring river widths. The cross-sectional measure is conducted on the river-reach (graph edge) basis, where orthogonals to centerlines are determined by the bounding geometry. In our experiments, the output results of GrabRiver are consistent with the reference river widths (R2 = 0.98 in the mean width validation and R2 = 0.91 in the transient width validation). Despite that GrabRiver is a promising method, random uncertainty in water mapping is identified as the major source of width measurement errors. The outputs of GrabRiver will be applicable in fluvial analysis and satellite-derived discharge estimates. [ABSTRACT FROM AUTHOR]

Steep-slope switching is effective to reduce the required energy for switching, however, at least 60 mV of gate voltage is required to modulate the current by an order of magnitude at room temperature. In this article, a numerical study of cold-source transistors based on chiral graphene nanoribbons (CGNRs) is presented. In cold-source transistors, the high-energy electrons are filtered out to break the room-temperature limitation, which can be realized by using CGNRs with narrow density of states (DOS) distribution. Our numerical results indicate that CGNR transistors can achieve sub-60 mv/decade subthreshold swing and similar ON-state current to conventional transistor. Moreover, the effect of the DOS distribution of CGNRs on the transport characteristics is investigated. This work provides a potential option for low-power electron devices and provides guidance for the design of cold-source transistors. [ABSTRACT FROM AUTHOR]

This paper presents an algorithm for the retrieval of ozone profiles from the Cross-track Infrared Sounder (CrIS) aboard the Suomi National Polar-orbiting Partnership satellite using a nonlinear optimal estimation method. The issue of channel selection is discussed. Based on a sensitivity analysis, we selected a spectral range of 990–1070 cm−1 for the ozone profile retrievals. Compared with ERA-Interim ozone profiles and eigenvector regression method profiles, the ozone climatology profile is better able to construct the a priori state. The retrieved CrIS profiles are in good agreement with smoothed high-vertical-resolution ozonesonde profiles. An analysis of the information content of the CrIS retrievals demonstrates that the CrIS measurements can provide useful information for capturing the spatial and temporal variations in ozone and are insensitive to ozone below 400 hPa. An error analysis revealed that smoothing error represents the main error source for the retrieved CrIS ozone profiles. [ABSTRACT FROM AUTHOR]

Scene classification plays an important role in the interpretation of remotely sensed high-resolution imagery. However, the performance of scene classification strongly relies on the discriminative power of feature representation, which is generally hand-engineered and requires a huge amount of domain-expert knowledge as well as time-consuming hand tuning. Recently, unsupervised feature learning (UFL) provides an alternative way to automatically learn discriminative feature representation from images. However, the performances achieved by conventional UFL methods are not comparable to the state-of-the-art, mainly due to the neglect of locally substantial image structures. This paper presents an improved UFL algorithm based on spectral clustering, named UFL-SC, which cannot only adaptively learn good local feature representations but also discover intrinsic structures of local image patches. In contrast to the standard UFL pipeline, UFL-SC first maps the original image patches into a low-dimensional and intrinsic feature space by linear manifold analysis techniques, and then learns a dictionary (e.g., using K-means clustering) on the patch manifold for feature encoding. To generate a feature representation for each local patch, an explicit parameterized feature encoding method, i.e., triangle encoding, is applied with the learned dictionary on the same patch manifold. The holistic feature representation of image scenes is finally obtained by building a bag-of-visual-words (BOW) model of the encoded local features. Experiments demonstrate that the proposed UFL-SC algorithm can extract efficient local features for image scenes and show comparable performance to the state-of-the-art approach on open scene classification benchmark. [ABSTRACT FROM PUBLISHER]