Search

Despite remarkable progress in knowledge transfer across visual and textual domains, extending these achievements to indoor localization, particularly for learning transferable representations among Received Signal Strength (RSS) fingerprint datasets, remains a challenge. This is due to inherent discrepancies among these RSS datasets, largely including variations in building structure, the input number and disposition of WiFi anchors. Accordingly, specialized networks, which were deprived of the ability to discern transferable representations, readily incorporate environment-sensitive clues into the learning process, hence limiting their potential when applied to specific RSS datasets. In this work, we propose a plug-and-play (PnP) framework of knowledge transfer, facilitating the exploitation of transferable representations for specialized networks directly on target RSS datasets through two main phases. Initially, we design an Expert Training phase, which features multiple surrogate generative teachers, all serving as a global adapter that homogenizes the input disparities among independent source RSS datasets while preserving their unique characteristics. In a subsequent Expert Distilling phase, we continue introducing a triplet of underlying constraints that requires minimizing the differences in essential knowledge between the specialized network and surrogate teachers through refining its representation learning on the target dataset. This process implicitly fosters a representational alignment in such a way that is less sensitive to specific environmental dynamics. Extensive experiments conducted on three benchmark WiFi RSS fingerprint datasets underscore the effectiveness of the framework that significantly exerts the full potential of specialized networks in localization., Comment: Accepted in the 1st round at WACV 2025 (Algorithm Track)

We propose iMoT, an innovative Transformer-based inertial odometry method that retrieves cross-modal information from motion and rotation modalities for accurate positional estimation. Unlike prior work, during the encoding of the motion context, we introduce Progressive Series Decoupler at the beginning of each encoder layer to stand out critical motion events inherent in acceleration and angular velocity signals. To better aggregate cross-modal interactions, we present Adaptive Positional Encoding, which dynamically modifies positional embeddings for temporal discrepancies between different modalities. During decoding, we introduce a small set of learnable query motion particles as priors to model motion uncertainties within velocity segments. Each query motion particle is intended to draw cross-modal features dedicated to a specific motion mode, all taken together allowing the model to refine its understanding of motion dynamics effectively. Lastly, we design a dynamic scoring mechanism to stabilize iMoT's optimization by considering all aligned motion particles at the final decoding step, ensuring robust and accurate velocity segment estimation. Extensive evaluations on various inertial datasets demonstrate that iMoT significantly outperforms state-of-the-art methods in delivering superior robustness and accuracy in trajectory reconstruction., Comment: Accepted as technical research paper in 39th AAAI Conference on Artificial Intelligence, 2025 (AAAI 2025)

Recent advancements in non-invasive health monitoring technologies underscore the potential of mm-Wave Frequency-Modulated Continuous Wave (FMCW) radar in real-time vital sign detection. This paper introduces a novel dataset, the first of its kind, derived from mm-Wave FMCW radar, meticulously capturing heart rate and respiratory rate under various conditions. Comprising data from ten participants, including scenarios with elevated heart rates and participants with diverse physiological profiles such as asthma and meditation practitioners, this dataset is validated against the Polar H10 sensor, ensuring its reliability for scientific research. This dataset can offer a significant resource for developing and testing algorithms aimed at non-invasive health monitoring, promising to facilitate advancements in remote health monitoring technologies.

This study presents a non-invasive method using thermal imaging to estimate heart and respiration rates in calves, avoiding the stress from wearables. Using Kernelised Correlation Filters (KCF) for movement tracking and advanced signal processing, we targeted one ROI for respiration and four for heart rate based on their thermal correlation. Achieving Mean Absolute Percentage Errors (MAPE) of 3.08% for respiration and 3.15% for heart rate validates the efficacy of thermal imaging in vital signs monitoring, offering a practical, less intrusive tool for Precision Livestock Farming (PLF), improving animal welfare and management.

Recognizing animal activities holds a crucial role in monitoring animals' health and well-being. Additionally, a considerable audience is keen on monitoring their pets' well-being and health status. Insight into animals' habitual activities and patterns not only aids veterinarians in accurate diagnoses but also offers pet owners early alerts. Traditional methods of tracking animal behavior involve wearable sensors like IMU sensors, collars, or cameras. Nevertheless, concerns, including privacy, robustness, and animal discomfort persist. In this study, radar technology, a noninvasive remote sensing technology widely employed in human health monitoring, is explored for AAR. Radar enables fine motion analysis through Microdoppler spectrograms. Utilizing an off-the-shelf FMCW mm-wave radar, we gather data from five distinct activities and postures. Merging radar technology with Machine Learning and Deep Learning algorithms helps distinguish diverse pet activities and postures. Specific challenges in AAR, such as random movements, being uncontrollable, noise, and small animal size, make radar adoption for animal monitoring complex. In this study, RayPet unveils different challenges and solutions regarding monitoring small animals. To overcome the challenges, different signal processing steps are devised and implemented, tailored for animals. We use four types of classifiers and achieve an accuracy rate of 89%. This progress marks an important step in using radar technology to observe and comprehend activities and postures in pets in particular and in animals in general, contributing to our knowledge of animal well-being and behavior analysis.

Field-Programmable Gate Array (FPGA)-based Software-Defined Radio (SDR) is well-suited for experimenting with advanced wireless communication systems, as it allows to alter the architecture promptly while obtaining high performance. However, programming the FPGA using a Hardware Description Language (HDL) is a time-consuming task for FPGA developers and difficult for software developers, which limits the potential of SDR. High-Level Synthesis (HLS) tools aid the designers by allowing them to program on a higher layer of abstraction. However, if not carefully designed, it may lead to a degradation in computing performance or significant increase in resource utilization. This work shows that it is feasible to design modern Orthogonal Frequency Division Multiplex (OFDM) baseband processing modules like channel estimation and equalization using HLS without sacrificing performance and to integrate them in an HDL design to form a fully-operational FPGA-based Wi-Fi (IEEE 802.11a/g/n) transceiver. Starting from no HLS experience, a design with minor overhead in terms of latency and resource utilization as compared to the HDL approach was created in less than one month. We show the readability of the sequential logic as coded in HLS, and discuss the lessons learned from the approach taken and the benefits it brings for further design and experimentation. The FPGA design generated by HLS was verified to be bit-true with its MATLAB implementation in simulation. Furthermore, we show its practical performance when deployed on a System-on-Chip (SoC)-based SDR using a professional wireless connectivity tester., Comment: 7 pages, extended version of poster accepted at FCCM 2023

We present a computational framework for two-scale asymptotic homogenization to determine the intrinsic magnetic permeability of composites. To this end, considering linear magnetostatics, both vector and scalar potential formulations are used. Our homogenization algorithm for solving the cell problem is based on the displacement method presented in Lukkassen et al. 1995, Composites Engineering, 5(5), 519-531. We propose the use of the meridional eccentricity of the permeability tensor ellipsoid as an anisotropy index quantifying the degree of directionality in the linear magnetic response. As application problems, 2D regular and random microstructures with overlapping and nonoverlapping monodisperse disks, all of which are periodic, are considered. We show that, for the vanishing corrector function, the derived effective magnetic permeability tensor gives the (lower) Reuss and (upper) Voigt bounds with the vector and scalar potential formulations, respectively. Our results with periodic boundary conditions show an excellent agreement with analytical solutions for regular composites, whereas, for random heterogeneous materials, their convergence with volume element size is fast. Predictions for material systems with monodisperse overlapping disks for a given inclusion volume fraction provide the highest magnetic permeability with the most increased inclusion interaction. In contrast, the disk arrangements in regular square lattices result in the lowest magnetic permeability and inadequate inclusion interaction. Such differences are beyond the reach of the isotropic effective medium theories, which use only the phase volume fraction and shape as mere statistical microstructural descriptors.

Software Defined Radio (SDR) platforms are valuable for research and development activities or high-end systems that demand real-time adaptable wireless protocols. While low latency can be achieved using the dedicated digital processing unit of a state-of-the-art SDR platform, its Radio Frequency (RF) front-end often poses a limitation in terms of turnaround time (TT), the time needed for switching from the receiving to the transmitting mode (or vice versa). Zero Intermediate Frequency (ZIF) transceivers are favorable for SDR, but suffer from self-interference even if the device is not currently transmitting. The strict MAC-layer requirements of Time Division Duplex (TDD) protocols like Wi-Fi cannot be achieved using configurable ZIF transceivers without having to compromise receiver sensitivity. Using a novel approach, we show that the TT using the AD9361 RF front-end can be as low as 640 ns, while the self-interference is at the same level as achieved by the conventional TDD mode, which has a TT of at least 55 {\mu}s. As compared to Frequency Division Duplex (FDD) mode, a decrease of receiver noise floor by about 13 dB in the 2.4 GHz band and by about 4.5 dB in the 5 GHz band is achieved., Comment: Accepted by IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks (WoWMoM 2022) as Work-In-Progress paper

The HL7 standard is widely used to exchange medical information electronically. As a part of the standard, HL7 defines scalar communication data types like physical quantity, point in time and concept descriptor but also complex types such as interval types, collection types and probabilistic types. Typical HL7 applications will store their communications in a database, resulting in a translation from HL7 concepts and types into database types. Since the data types were not designed to be implemented in a relational database server, this transition is cumbersome and fraught with programmer error. The purpose of this paper is two fold. First we analyze the HL7 version 3 data type definitions and define a number of conditions that must be met, for the data type to be suitable for implementation in a relational database. As a result of this analysis we describe a number of possible improvements in the HL7 specification. Second we describe an implementation in the PostgreSQL database server and show that the database server can effectively execute scientific calculations with units of measure, supports a large number of operations on time points and intervals, and can perform operations that are akin to a medical terminology server. Experiments on synthetic data show that the user defined types perform better than an implementation that uses only standard data types from the database server., Comment: 12 pages, 9 figures, 6 tables

This paper is about Kripke structures that are inside a relational database and queried with a modal language. At first the modal language that is used is introduced, followed by a definition of the database and relational algebra. Based on these definitions two things are presented: a mapping from components of the modal structure to a relational database schema and instance, and a translation from queries in the modal language to relational algebra queries.