In this paper, we introduce a framework for multi-class novelty detection using structural vibration signals. Structural vibration-based person identification is a promising soft-biometric approach with potential applications in elderly care and access control. However, current research faces two key challenges. The first challenge is the lack of large-scale datasets necessary for thorough evaluation in structural vibration gait recognition. To address this, we created a new dataset with recordings from fifty individuals. The second challenge lies in the limited exploration of deep learning methods for large-scale multi-class novelty detection in structural vibration data. To fill this gap, we propose the energy-shifted contrastive loss function, specifically designed for this task. Our results demonstrate that the proposed framework achieves 96.57% accuracy in multi-class classification. For novelty detection, it achieves an Receiver Operating Characteristic-Area Under the Curve (ROC-AUC) score of 89.15% for single footsteps, which improves to 93.83% with five consecutive footsteps.

This article introduces a large-scale, nonintrusive person identification (PrID) framework using footstep-induced structural vibration signals. The increasing adoption of structural vibration analysis for PrID comes from its inherent nonintrusiveness and privacy preserving characteristics. However, the existing methodologies are often constrained by the scarcity of extensive datasets, both in terms of the number of subjects and the temporal length of individual recordings, and frequently rely on supervised learning paradigms coupled with manual feature engineering. Consequently, the generalization capabilities of these approaches to broader populations are typically limited. To address these limitations, we have curated a comprehensive dataset of structural vibration signals acquired from 100 individuals. In addition, we have developed an unsupervised event detection method using the features based on time, frequency, and wavelet analysis. Furthermore, we have developed DeepStep, a residual attention-based framework specifically designed for efficient feature extraction and classification of structural vibration signals. Experimental evaluation on our curated dataset demonstrates that the proposed approach achieves a Rank-1 accuracy of approximately 92% and a Rank-5 accuracy of approximately 96%.

Structural vibration-based gait recognition has emerged as a promising soft-biometric modality, particularly for privacy-sensitive monitoring and access control. Despite its potential, current research is largely limited to proof-of-concept studies that rely on hand-crafted features, with minimal exploration of deep learning methodologies. This gap reduces the potential for integrating structural vibration-based gait recognition with existing modalities, such as camera-based systems. In this study, we propose a multi-modal gait recognition system that integrates both vision and structural vibration modalities. We address two key challenges: (a) lack of studies exploring outdoor gait recognition using both vision and structural vibration, and (b) absence of a multi-modal training scheme that combines these two modalities. To tackle the first challenge, we curated a dataset comprising five minutes of walking data from ten individuals captured simultaneously by two cameras and a geophone sensor. To address the second challenge, we developed a joint training framework that uses data from both modalities. Our methods achieve an accuracy of 96.03% (+/- 1.12) using structural vibration signals alone, and this improves to 98.27% (+/- 0.06) when both modalities are combined.

Gait recognition based on structural vibration signals is an emerging area in soft biometrics and healthcare. It is valued for its privacy-preserving features, making it ideal for continuous monitoring in healthcare environments. In this work, we propose a method for simultaneous person identification and gait-abnormality detection using structural vibration signals. We have experimented on a dataset of eight people. The system uses shared feature extraction layers with separate branches for identification and abnormality detection. Our framework achieves a person identification accuracy of similar to 89.00% and similar to 90.00% accuracy in detecting abnormality in gait patterns.

In correlation-based processing, sparse arrays offer the capacity to resolve a greater number of uncorrelated sources than physical sensors due to the considerable breadth of their difference coarrays, originating from variations in the locations of elements. Consequently, there is significant interest in devising sparse arrays with sizable difference coarrays and expanding the analysis to encompass additional array characteristics like symmetry, resilience, and cost-effective engineering. We present a scalable and systematic methodology for designing large sparse arrays. Considering several attributes and factors, we can address Fractal arrays that were used for low-side lobe antenna array designing and have very low degrees of freedom; hence, sparsity is introduced to design a hole-free difference coarray which not only increases the number of degrees of freedom in fractal arrays but also aids in better beamforming applications and enhanced DoA results due to regularization in coarrays. We develop an innovative sparse fractal array to enhance the accuracy of DoA estimation for predicting a maximum number of uncorrelated sources with a minimum possible actual sensors. First, the 1D sparse fractal array is constructed and then it is extended to a 2D sparse fractal array for both azimuth and elevation angle estimation. Comprehensive robustness analysis is conducted on the proposed sparse fractal array, encompassing one-dimensional (1D) and two-dimensional (2D) configurations, in response to sensor failures. RMSE analysis shows that the proposed 1D and 2D arrays possess the minimum error when used for direction estimation.

A novel strategy for improving signal quality in the DWDM based networks is proposed in this article. Signal quality enhancement in an optical network is a non-trivial task which depends on the operating points and characteristics of the key optical devices like in-line amplifier, pre-amplifier and wavelength selective switch. Commercial-grade devices are used for creating an optical network in CDOT optical testbed and evolution of the strategy for signal quality enhancement. The evolved strategy is validated and compared with the similar approaches described in the literature based on the metrics of optical signal quality. It is observed that the proposed strategy reduces the standard deviation of received channel powers by around $\mathbf{6 5 \%}$ (relative to the baseline strategy) while also improving OSNR. The enhancement in signal quality is validated in multiple spectrum load scenarios. It is also experimentally verified that the proposed strategy is resilient to dynamic changes in the network.

We introduce "GajGamini:" a novel method for detecting elephant movement by analyzing ground vibrations recorded using seismic sensors. This method is based on the principle that ground vibrations from elephants are distinct from those caused by humans and background noise. In this letter, we address two main challenges. First, there was a lack of studies with extensive data on vibrations from Indian elephants and humans. To address this, we recorded 3 h of elephant movements and 2 h of human movements using seismic sensors. Second, there was a need for a dedicated architecture for the real-time classification of seismic vibrations from elephants, humans, and background noise. To overcome this, we propose a convolutional neural network (CNN)-based model named "GajGamini" that achieves a prediction accuracy of similar to 98.03% with only 3 s of computational runtime for every 10 s of recorded data. GajGamini represents a significant advancement in wildlife monitoring, particularly for elephant conservation. It offers a noninvasive way to track elephant movements, enhancing the effectiveness of wildlife management strategies.

Contour tracing is a critical technique in image analysis and computer vision, with applications in medical imaging, big data analytics, machine learning, and robotics. We introduce a novel hardware accelerator based on the adapted and segmented (AnS) vertex following (VF) and run-data-based-following (RDBF) families of fast contour tracing algorithms implemented on the Zynq-7000 field-programmable gate array (FPGA) platform. Our algorithmic implementation utilizing a mesh-interconnected multiprocessor architecture is at least 55x faster than the existing implementations. With input-output overheads, it is up to 12.5x faster. Our hardware accelerator for contour tracing is benchmarked on mesh-interconnected hardware, all three families of contour tracing algorithms, and a random image from the Imagenet database. Our implementation is, thus, faster for FPGA, application-specific integrated circuit (ASIC), graphics processing unit (GPU), and supercomputer hardware in comparison to the central processing unit (CPU)-GPU collaborative approach and offers a better solution for those systems where the input-output overheads can be minimized, such as parallel processing arrays and mesh-connected sensor networks.

The massive data demand requires content distribution networks (CDNs) to use evolving techniques for efficient content distribution with guaranteed quality of service (QoS). The distributed fog-based CDN model, with optimal fog node placements, is a suggested aproach by researchers to meet this demand. While many studies have focused on improving QoS by optimizing fog node placement, they have rarely considered the impact on content distribution, affected by placement, usage changes, and delivery rates. Therefore, the practical approach to fog node placement for CDN services must examine its impact on content distribution. Further, current research on fog-based CDN lacks formal methods to address key challenges: R1) strategic placement of fog nodes to process end-user requests; R2) construction of a content distribution path with guaranteed QoS; R3) cost minimization of building a fog-based CDN model. We construct this as a joint optimization problem by considering four parameters: geographical regions, open public Wi-Fi access points (OPWAPs) locations, QoS, and cost to achieve research objectives R1–R3. As a solution, we propose a dual-step framework. First, a heuristic for optimal fog node placement based on geographic regions and OPWAP locations is proposed. Second, we propose two algorithms, Greedy Performance-based Node Selection (GPDS) and Greedy Fog Node Selection algorithm (GFNSA), for selecting fog nodes, minimizing the cost of building a fog-based CDN while achieving optimal content distribution paths. The results demonstrate that the proposed methods outperform the baseline techniques and provide near-optimal solutions to the problem.

Systems that use Deep Learning (DL) models extensively utilize cloud computing for inference tasks in various domains such as traffic monitoring, healthcare, and IoT. However, applications like autonomous vehicles, surveillance systems, and spacecraft are transitioning towards edge computing due to band-width limitations, transmission delays, and network connectivity issues. Edge computing mitigates these challenges by reducing latency through local data and model processing on the device. Implementing Deep Neural Networks (DNNs) on edge devices faces resource constraints, such as limited memory, computing power, etc. DNNs employ 32-bit floating-point precision for accuracy, leading to inflated model sizes. Quantization offers a solution by converting high-precision floating-point (FP) values to lower-precision or integer (INT) values, focusing on throughput and improving latency. This paper presents a comparative study of the accuracy and performance of 64-bit, 32-bit, and 16-bit floating-point instructions, along with 8-bit integer instructions, using Post Training Quantization (PTQ) and Quantization-Aware Training (QAT), on multiple Nets including CustomNets, which was inferenced on a GPU as well as a Xilinx Deep Processing Unit (DPU). The models were evaluated on a sample of the EuroSat Remote Sensing dataset. Quantizing models to FP16 and INT8 resulted in $2-3 x$ and $4 x$ faster inferencing, respectively, with a negligible decrease in accuracy of $1-4 \%$. FP64 exhibited a 2 $3 x$ decrease in speed but a slight accuracy improvement ($2 \%$). On the DPU, models showed minimal accuracy degradation of about $1 \%$. Overall, model size decreased by a constant $2 x$ and 4x from FP32 to FP16 and INT8, respectively, while increasing by $2 x$ for FP64. This reduction in size, with negligible loss in accuracy enables onboard storage along withfaster and accurate inferencing on resource constraint systems.

Efficient content distribution to end-users poses a significant challenge in content distribution networks (CDNs). Traditional CDNs rely on cloud-based architectures, which may not be optimal for delivering content in densely populated areas due to increased network latency and bandwidth limitations. These problems can be mitigated by adopting fog/edge computing, which deploys computing nodes near end-users to reduce latency and improve content delivery. Although factors such as deployment and distribution are often considered separately, there are relatively few studies on how node deployment affects content distribution. Moreover, current research on fog-based content distribution networks (fog-based CDN) does not often address formal methods for key challenges such as (R1) optimal fog node placement; (R2) providing efficient content distribution to end-users; (R3) minimizing the overall fog-based CDN cost. Therefore, we propose an algorithm to jointly optimize the placement of fog nodes and the content distribution, called the Joint Optimization Algorithm for the Fog-node Placement and Content distribution (JFnP-CDA). The algorithm uses a two-step procedure including clustering and Voronoi method, and nonlinear programming to optimize R1, R2, and R3. We consider four parameters for this: a given geographical region, locations of open public Wi-Fi access points (OPWAPs) in that region, quality of service (QoS, with delivery in delay as the measure), and cost to generate optimal service sub-regions (and, thereby, distribute content to edge-network hot spots). We evaluated the effectiveness of the proposal by implementing real-world OPWAP data, and the results show that the algorithm JFnP-CDA outperforms the baseline methods.

Developing a novel category of 2-D sparse arrays to improve information resolution with a minimum of possible actual sensors has proven to be a notable and persistent demanding objective. This study introduces a new layout for planar sparse arrays to estimate signal arrival angles in both azimuth and elevation dimensions. By introducing this new category of two-dimensional arrays, we strive to improve existing planar geometries. Specifically, our method’s principle involves strategically positioning sensors to maximize spatial diversity and enhance the array’s ability to resolve incoming signals from different angles. This optimization results in a higher DoF compared to traditional planar arrays, offering improved performance in signal processing tasks such as direction finding and beamforming. Furthermore, the intended design of the array aims to reduce the occurrence of gaps within the difference coarray and make it hole-free to enhance its suitability for DoA estimation algorithms. To assess the effectiveness of the suggested array, numerical simulations were conducted, demonstrating the achieved RMSE (Root Mean Square Error) value in the proposed RCPA is the least, i.e., less than 0.1 % , as compared to other existing planar arrays. The reduced RMSE achieved with the proposed RCPA array could significantly enhance the accuracy of existing applications such as radar and sonar systems, leading to improved target detection, tracking, and localization capabilities. The detailed explanation of our approach and the rationale behind its improved DoF serves to provide readers with a comprehensive understanding and motivation for adopting this innovative array design.

In biometrics, footstep-based seismic signals for person identification have become increasingly popular. Person identification is a crucial aspect because it serves as the foundation for personalized services like elderly monitoring, tailored access control, etc. However, these systems may not have enough data for specific individuals, especially those with limited mobility. This lack of data can be a significant challenge, making achieving accurate and reliable results difficult. We used Generative Adversarial Networks (GAN) to develop a novel technique for generating a footstep-based seismic signal of a person. In this letter, we have addressed two challenges. The first one involves generating synthetic footstep signals that closely mimic the significant statistical properties of an individual's footstep signal. The second challenge is filtering the generated data to achieve high classification accuracy. To overcome the first challenge, we have developed a Wasserstein distance-based One-Dimensional Generative Adversarial Network(1DGAN) with gradient penalty, to generate synthetic footstep data of a person. To address the second challenge we formulated a screening process based on statistical error metrics. Our experimentation on ten people achieved an overall test identification accuracy of 98.44%. Furthermore, it achieved an average identification accuracy of 91.23% from just ten footsteps when tested with unseen real data.

Metal matrix composites have attracted extensive attention from both the research and industrial perspective. In this study, we prepared aluminum-reduced graphene oxide (Al-rGO) composites with enhanced thermal conductivity in an easy single-step process. Pristine Al shows a thermal conductivity of 175 Wm(-1)K(-1) (standard deviation <5%), which increases to 293 Wm(-1)K(-1) for an Al-rGO composite with 1% rGO. Analysis of theoretical models shows that a higher percentage of rGO inside the Al matrix creates a continuous network resulting in more available phase space through which heat carrier phonons travel with less scattering, and hence thermal conductivity of the composite increases. Furthermore, Al-rGO composites show an & SIM;5% increase in microhardness compared with pristine Al. The electrical resistivity of the composite is comparable to that of pristine Al for a narrow weight percentage of rGO, whereas a 70% enhancement in the thermal conductivity of the composite is observed for the same weight percentage range, suggesting possibilities for exploiting both high electrical and thermal conductivities for various applications.

The manufacturing and characterization of freeform optical surfaces are influenced by their high sensitivity to misalignments. In this work, the computational sampling moiré technique combined with phase extraction is developed for the precise alignment of freeform optics during fabrication and in metrology applications. This novel, to the best of our knowledge, technique achieves near-interferometry-level precision in a simple and compact configuration. This robust technology can be applied to industrial manufacturing platforms (such as diamond turning machines, lithography, and other micro-nano-machining techniques) as well as their metrology equipment. In a demonstration of computational data processing and precision alignment using this method, iterative manufacturing of freeform optical surfaces with a final-form accuracy of about 180 nm was accomplished.

Torus networks connect processing elements in network-on-chips for several high-performance computing applications. Though square torus networks are the most popular, Rectangular Torus (RT) networks are also used where the number of multiprocessors is constrained. A twist in the rectangular torus topology can be used to reduce the average inter-node distance of the network at the expense of network symmetricity. The Average Inter-node (Shortest) Distance property of a Rectangular Twisted Torus (RTT) Network is an important parameter to decide the twist to be applied to the RT network. Also, the regular mapping of an RTT network on a VLSI chip can take up multiple layers during implementation, which further limits the use of RTT in an application. The Average Inter-node Distance (AID), as a function of the twist, is examined and presented for a family of Rectangular Twisted Torus (RTT) networks. The two-dimensional mapping, using only two routing layers on a VLSI chip, of a subset of these RTT networks is also presented. The AID, as a function of the twist, can be used to select the RTT network to be used for an NoC application. The two-layer mapping of that RTT can be further used for implementation on an ASIC or FPGA.

Received signal strength (RSS)-based localization has been popular as it can achieve reasonable accuracy without the need for additional hardware. In this work, we consider RSS-based cooperative and non-cooperative localization when the transmit power of the nodes is unknown. The maximum likelihood (ML) formulation of the RSS-based localization is non-linear, non-convex, and discontinuous and cannot be solved using conventional techniques. Firstly, we linearize the relation between pairwise distance between nodes and received power using a least squares-based linearization technique. Next, we propose two RSS-based localization techniques, referred to as SDP-URSS and SDP-RSS, that convert the ML into a constrained convex optimization problem using the proposed linearization technique and semidefinite relaxation. SDP-URSS assumes the transmit powers are unknown and estimates them along with the location of the nodes, whereas SDP-RSS uses the transmit power information to improve the location estimates. Extensive performance evaluation of the proposed technique considering various performance metrics under non-cooperative and cooperative scenarios demonstrates the superiority of the proposed localization techniques over the existing methods.

The term fractal refers to the fractional dimensions that have recursive nature and when clubbed with the properties of sparse arrays leads to the generation of a novel array called a sparse fractal array. In this paper, we extend our research to the 2D domain by introducing planar sparse arrays which generate hole-free difference coarray and have OpN 2 q elements just like the OBA but here in the new closed box form, with the additional property of fractal arrays along with sparseness. To estimate azimuth and elevation angle we have designed planar sparse fractal arrays using nested arrays and coprime arrays as the fundamental basic generating array which helps in achieving a high degree of freedom which makes it useful for DOA estimation. Simulations show that the proposed planar arrays have the better estimation performance when compared with existing planar arrays like URA, OBA, and CPA.

The term fractal refers to the fractional dimensions that have recursive nature and exhibit better array factor properties. In this article, we present a new class of sparse arrays where the recursive nature of a fractals is used in designing an antenna array called a sparse fractal array by combining sparsity properties of the various sparse array and the recursive nature of the fractal array. The most important property of the proposed array is the hole-free difference coarray which makes it a good choice for DOA estimation. Upon carrying out the robustness analysis of the proposed geometry in sensor failure environment it is found that the proposed array is robust enough as compared to already existing sparse arrays which shows highest fragility value. Direction of arrival estimation of the uncorrelated sources using the proposed geometry showed the RMSE(root mean square error) value of the order of 10−3 which proves it to be a good choice for the application of DOA estimation algorithms like MUSIC, ESPIRIT etc.

The ability to determine the location of a sensor node in a wireless sensor network is essential in many applications. We propose a received signal strength (RSS) based cooperative localization technique. The target nodes, whose locations are unknown during the deployment phase, communicate with the anchor nodes (whose locations are known) and the neighboring target nodes to localize themselves. In the proposed technique the target nodes behave as pseudo anchor nodes after they have localized themselves. It is observed that in such cooperative techniques the localization error propagates from one iteration to the next. We mitigate this error propagation by eliminating the network edge nodes from updating their location information after the first iteration. We propose two distributed edge node detection techniques: quadrant method (Quad) and geometric dilution of precision (GDOP) based method. Extensive simulations have been carried out, on uniform anchor and non-uniform network topologies, to analyze the performance of the proposed edge detection techniques and the iterative localization algorithm. The proposed localization algorithm is compared with state of the art localization techniques using simulations as well as experiments, and is shown to achieve improved localization accuracy. The Cramer-Rao lower bound (CRLB) is also derived and used as a performance benchmark.