Distributed Acoustic Sensing (DAS) enables large-scale monitoring through optical fibers, but its high dimensionality and complex spatio-temporal patterns make event classification demanding. Existing deep learning approaches–CNNs, recurrent models, and Transformer variants–either fail to capture long-range dependencies or require processing raw DAS matrices at prohibitive cost. We propose DAStatFormer, a hybrid multibranch Transformer that combines compact multidomain statistical features with Gated Transformer Networks. Instead of raw signals, we extract 24 ANOVA-selected attributes per channel from the temporal, waveform, and spectral domains, reducing data size by orders of magnitude while preserving discriminative information. Each domain is processed via dedicated step-wise and channel-wise attention branches, fused by an adaptive gating mechanism. Experiments on the open ϕ -OTDR benchmark and a real-scenario DAS dataset show that DAStatFormer achieves up to 99.4 https://github.com/MichelD-git/DAStatFormer

Passive infrared (PIR) sensors provide a low-cost, unobtrusive, and privacy-preserving approach for continuously monitoring daily activity in older adults. This study investigated whether indoor mobility features derived from PIR sensors could discriminate levels of health-related quality of life (HRQoL) in community-dwelling older adults living alone. Mobility variables were extracted from three months of PIR sensor recordings and aggregated at the participant level for 40 individuals, who were classified into high- and low-HRQoL groups according to the EQ-5D index. A nested stratified five-fold cross-validation framework was implemented, incorporating RandomOverSampler exclusively within the training folds to address class imbalance while preserving the original distribution of the outer test folds. Three machine learning classifiers—Support Vector Machine (SVM), Random Forest, and K-Nearest Neighbors (KNN)—were evaluated using accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUC). The SVM achieved the best overall performance, with an accuracy of 0.825±0.143, precision of 0.860±0.080, recall of 0.900±0.149, F1-score of 0.876±0.102, and AUC of 0.937±0.069. Random Forest achieved a comparable AUC of 0.933±0.109, whereas KNN showed lower overall performance and greater variability across the outer folds. Aggregated out-of-fold predictions further provided class-specific performance estimates while preserving the original participant distribution, confirming that model evaluation was conducted exclusively on non-oversampled test data. Overall, the findings support the feasibility of combining PIR-derived mobility features with interpretable machine learning models to investigate HRQoL in older adults. These results should be considered proof-of-feasibility and warrant validation in larger, independent, and more diverse cohorts.

Deep learning plays a vital role in wearable physiological monitoring, and smart-textile garments that embed soft sensors directly into fabric have extended this capability to home-based healthcare, rehabilitation, sleep monitoring, and occupational safety. Nevertheless, smart-textile sensing faces a fundamental reliability limitation that conventional wearable pipelines do not: an abnormal measurement collected from a sensorized garment may originate from a genuine physiological event, from degradation of the textile sensor itself, or from a transient garment–body interaction artifact. Textile sensors are known to exhibit sensitivity drift, hysteresis, baseline shift, contact loss, and humidity-dependent behaviour under repeated strain, laundering, and long-term use; consequently, an unusual waveform is intrinsically ambiguous. This work therefore addresses a human-centric reliability problem in wearable health intelligence: when a smart garment reports an abnormal signal, the system must decide whether the user needs physiological attention or whether the garment itself requires recalibration, repositioning, or replacement, since each conclusion leads the user, caregiver, clinician, or textile engineer to a different action. To address this challenge, we design HCLM-SmartTex, an explainable and frugal framework grounded in Human-Centered Learning Mechanics (HCLM) that separates physiological anomalies from smart-textile measurement failures rather than conflating them. Crucially, the framework is optimised not for raw accuracy but against an explicit human-centered decision cost that penalises missed physiological events, textile-induced false alarms, and detection delay, so that the system behaves as a risk-aware decision-support tool rather than a mere signal classifier. The framework combines two transparent mechanistic sub-scores, respectively summarising physiological evidence and textile-reliability evidence, with a lightweight late-fusion classifier. An active HCLM control layer then regulates the fused score through an entropygated mechanism that moves the decision toward calibrated mechanism-derived risk evidence whenever predictive uncertainty and textile-reliability ambiguity are active. Because the true origin of an abnormal reading is rarely observable in real recordings, we introduce a controlled synthetic smart-textile simulator that supplies the counterfactual ground truth this problem requires (identical physiological events with and without sensor faults, and identical textile faults under different physiological states), generating apnea, hypopnea, sensor detachment, textile drift, and garment-misplacement scenarios with known cause. The proposed framework was evaluated against rule-based thresholds, Isolation Forest, One-Class SVM, a lightweight logistic classifier, an accuracy-boosted ExtraTrees baseline, and a non-frugal accuracy-optimized ensemble. On the held-out subject split, the active HCLM-SmartTex detector achieved the best F1-score (0.9515), the lowest false-alarm window fraction (0.0267), and a mean detection delay of 0.50 windows, improving over the unregulated XAI-SmartTex hybrid, whose F1-score and false-alarm window fraction were 0.9289 and 0.0433, respectively. The active HCLM thermostat selected λmax = 0.30, modified 64% of evaluated windows by more than 0.01 score units, and changed five test decisions relative to the hybrid, confirming that the control layer acted as a measurable risk regulator rather than collapsing to the hybrid. The framework further preserved process-aware explanations, a family-separation accuracy of 0.9667, and a root-cause accuracy of 0.9667 on the held-out fold, covering the active classes present in that fold; full six-class validation remains to be assessed under repeated subject-wise splits. It retained an embedded-compatible footprint of only 201 parameters, 2.15 KB RAM, 6.79 KB flash, and approximately 0.55 μJ per inference in the notebook operating point. As a further contribution, we provide a concrete physical smart-garment validation protocol (minimal instrumentation, a staged acquisition plan, a two-layer physiological and textile-reliability label structure, and explicit synthetic-to-real transfer checks) that defines the route from these controlled mechanisms to real-world deployment. The study is thus intended as a mechanism-level, decision-oriented benchmark for reliable smart-textile monitoring, with clinical validation to follow through the provided protocol rather than being claimed here.

Distributed Acoustic Sensing (DAS) technology has emerged as a powerful tool for large-scale acoustic monitoring, transforming standard fiber optic cables into dense arrays of virtual microphones. When combined with artificial intelligence, particularly deep learning, DAS enables scalable and automated detection of acoustic events, making it a promising solution for whale monitoring across vast marine environments. In the field of marine bioacoustics, DAS provides significant advantages in terms of spatial coverage and robustness compared to traditional hydrophone arrays. This paper presents a novel deep learning-based approach to detect whale vocalizations from DAS data. The proposed method leverages a hybrid architecture combining Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) networks to extract both spatial and temporal features from waterfall diagrams derived from DAS recordings. A multi-stage preprocessing pipeline including frequency filtering and frequency-wavenumber (f-k) filtering is applied to enhance signal quality and isolate whale calls from background noise. Each DAS recording is treated as a spatio-temporal matrix, and sequences of such matrices are sequentially analyzed capturing temporal dependencies. Experimental evaluations on a public dataset from the Ocean Observatories Initiative (OOI) RCA North Cable demonstrate that the CNN-BiLSTM model outperforms CNN and CNN-LSTM baselines, achieving a F1-score of 96

This paper provides a comprehensive review of the integration of Spiking Neural Networks (SNNs) and Transformers, combining the energy efficiency of SNNs with the high performance of Transformer architectures. By leveraging the event-driven nature of SNNs and the powerful self-attention mechanism of Transformers, this fusion aims to address the challenges of high energy consumption in deep learning while improving task accuracy, especially for complex datasets. We introduce the core concepts of SNNs and Transformers, reviewing state-of-the-art methods for their combination, including hybrid architectures. The performance of each architecture is presented thanks to both static and neuromorphic datasets, highlighting their advantages and limitations. This review also discusses the challenges of integrating self-attention into spiking architectures and outlines future research directions to further enhance model performance and energy efficiency.

Distributed Acoustic Sensing (DAS) offers a scalable and resilient solution for real-time perimeter security by transforming optical fibers into dense arrays of vibration sensors. However, its high-dimensional, noisy, and spatio-temporally complex data make accurate event recognition challenging. Conventional deep learning models, such as Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, often struggle to capture long-range dependencies and rely heavily on handcrafted or single-domain features, limiting their effectiveness in complex intrusion scenarios. Designing compact and discriminative feature representations remains a critical challenge. To address these issues, we propose DASViT1D, a novel intrusion detection framework that integrates a one-dimensional Vision Transformer (ViT1D) with fused multi-domain features. Specifically, we extract and combine Mel-Frequency Cepstral Coefficients (MFCC), Redundant Discrete Fourier Transform (RDFT), and Discrete Wavelet Transform (DWT) features to capture complementary spectral and temporal characteristics. ViT1D then leverages self-attention to model long-range dependencies across time. Evaluated on a public.-OTDR DAS dataset with nine event classes, our approach achieves 93.5% accuracy, outperforming CNN-based baselines by nearly 8 points, with a false negative rate (FNR) of 0.009, a nuisance alarm rate (NAR) of 0.005, and fast inference (0.13 s/sample), demonstrating its suitability for real-time DAS-based intrusion detection.

Higher Education is constantly pushing to include soft skills in their curricula. An alternative could be to personalise their curricula so that students could better develop their soft skills. However, there are not many studies that investigate how to personalise the students' curricula based on their soft skills, especially if we consider multiple soft skills that need development. The aim of the article is to propose a recommender system framework based on soft skills in order to bridge the soft skills gap between the expected proficiency by employers and the actual proficiency of graduates. The approach is illustrated using real data from three cohorts of students that graduated in the years 2021, 2022 and 2023 at a French Higher Education Institution. We use a psychometric modelling approach to predict the soft skills proficiency of students within a genetic algorithm framework. We define three fitness functions and two aggregation methods, with which we can quantify the relevance of a set of courses across 10 different soft skills (e.g., Problem Solving, Leadership). The results show the recommendations to have, on average, a higher fitness than the actual courses taken by the students during the program. Moreover, there is significant evidence that the recommendations would allow the students to satisfy more of the soft skill targets compared to the courses the students actually took.

Gesture recognition is a tool to enable novel interactions with different techniques and applications, like Mixed Reality and Virtual Reality environments. With all the recent advancements in gesture recognition from skeletal data, it is still unclear how well state-of-the-art techniques perform in a scenario using precise motions with two hands. This paper presents the results of the SHREC 2024 contest organized to evaluate methods for their recognition of highly similar hand motions using the skeletal spatial coordinate data of both hands. The task is the recognition of 7 motion classes given their spatial coordinates in a frame-by-frame motion. The skeletal data has been captured using a Vicon system and pre-processed into a coordinate system using Blender and Vicon Shogun Post. We created a small, novel dataset with a high variety of durations in frames. This paper shows the results of the contest, showing the techniques created by the 5 research groups on this challenging task and comparing them to our baseline method.

The current epidemiology of the global population highlights an increasing number of aged individuals living with decreasing autonomy. Modern societies face the significant challenge of caring for persons with various disabilities, whether in institutions or at home. Health systems are ill-prepared in terms of staffing, economics, and public policies to manage the growing population of elderly individuals. This article discusses how technology can alleviate isolated lives, reduce hazards, and enhance human relationships, particularly the physician-patient relationship. Our interdisciplinary group focuses on developing innovative technologies to be implemented in the coming decades to improve the living conditions of elderly populations.

Recently the use of Unmanned Aerial Vehicles (UAVs), commonly referred to as drones, has exploded in several domains, leading to potential security issues. As such, estimating the exact position of those eventual malicious drones has become of crucial interest. However, computing an accurate and precise geolocation of these drones, especially in outdoor environments, remains challenging. This paper focuses on outdoor 3-dimensional (3D) drone geolocation techniques based on Radio Frequency (RF) signals. We first present a RF-based 3D drone geolocation dataset, and then apply and compare various geolocation techniques, ranging from geometrical-based to machine learning-based methods. We further propose a new hybrid method blending the two above categories of geolocation techniques, that achieves an average 3D error of the order of 11.7 meters within a search volume of about 520×560×115 m 3 , significantly below the one achieved with geometrical-based techniques, and with a reduced computational complexity compared to the regular machine-learning based techniques.

Over the last decade, Higher Education has focused more of its attention toward soft skills compared to traditional technical skills. Nevertheless, there are not many studies concerning the relation between the courses followed within an academic program and the development of soft skills. This work presents a practical approach to model the effects of courses on soft skills proficiency. Multiple Membership Ordinal Logistic Regression models are trained with real data from students of the 2021, 2022, and 2023 cohorts from the general engineering program in a French Higher Education institution. The results show that attending a postgraduate course in average increases the odds of being more proficient in terms of soft skills. Nonetheless, there is considerable variability in the individual effect of courses, which suggest there can be huge differences between courses. Moreover, the data also suggest great dispersion in the students' initial soft skill proficiency.

Falls are a significant health risk for older adults, and fall-related injuries are a leading cause of morbidity and mortality in this population. Elderlies are particularly vulnerable to falls due to age-related declines in mobility, balance, and muscle strength, as well as chronic medical conditions with use of certain medications. These injuries can range from minor bruises and scrapes to more severe like fractures, head trauma, or internal bleeding. To prevent falls in older adults, some solutions propose to ensure a safe living environment, others to maintain physical activity, and others to manage chronic medical conditions. This article presents the implementation and test of a system preventing hip fractures resulting from falls using a fall detection and prediction system designed to protect and alert individuals during falls.

The personalization of curriculum plays a pivotal role in supporting students in achieving their unique learning goals. In recent years, researchers have dedicated efforts to address the challenge of personalizing curriculum through diverse techniques and approaches. However, it is crucial to acknowledge the phenomenon of student forgetting, as individuals exhibit variations in limitations, backgrounds, and goals, as evidenced by studies in the field of learning sciences. This paper introduces the complex issue of fully individualizing a curriculum while considering the impact of student forgetting, presenting a comprehensive framework to tackle this problem. Moreover, we conduct two experiments to explore this issue, aiming to assess the difficulty of identifying relevant curricula within this context and uncover behavioral patterns associated with the problem. The findings from these experiments provide valuable prescriptive recommendations for educational stakeholders seeking to implement personalized approaches. Furthermore, we demonstrate the complexity of this problem, highlighting the need for our framework as an initial decision-making tool to address this challenging endeavor.

Wireless networks such as IEEE 802.11 networks are nowadays widely used. However, they are vulnerable to different forms of attacks such as the fake access point and the emission of intentional frequency sweeping jamming signals. In this paper, we present a Network-based Intrusion Detection System (NIDS) which uses a threshold-based approach to detect the two attacks on IEEE 802.11 networks. The NIDS analyses indicators derived from frame header information to determine the presence of the two attacks. We also discuss the limitations of this approach and give perspective for a new line of research based on a supervised learning model approach.

Smart homes are equipped with several sensor networks to keep an eye on both residents and their environment, to interpret the current situation and to react immediately. Handling large scale dataset of sensory events on real time to enable efficient interventions is challenging and very difficult. To deal with these data flows and challenges, traditional streaming data classification approaches can be boosted by use of incremental learning. In this paper, we presented two new Incremental SVM methods to improve the performance of SVM classification in the context of human activity recognition tasks. Two feature extraction methods elaborated by refining dependency sensor extraction feature and focusing on the last sensor event only have been suggested. On the other hand, a clustering based approach and a similarity based approach have been suggested to boost learning performance of the incremental SVM algorithms capitalizing on the relationship between data chunk and support vectors of previous chunk. We demonstrate through several simulations on two major publicly available data sets (Aruba and Tulum), the feasibility and improvements in learning and classification performances in real time achieved by our proposed methods over the state-of-the-art. For instance, we have shown that the introduced similarity-based incremental learning is 5 to 9 times faster than other methods in terms of training performances. Similarly, the introduced Last-state sensor feature method induces at least 5% improvement in terms of F1-score when using baseline SVM classifier.

Wireless networks are nowadays indispensable components of telecommunication infrastructures. They offer flexibility, mobility and rapid expansion of telecommunication infrastructures. In wireless networks, transmissions are unisolated and most commonly emitted using omnidirectional antennas. This makes wireless networks more vulnerable to some specific attacks as compared to wired networks. For instance, attacks such as fake access points, intentional jamming and deauthentication can be easily perpetrated against IEEE 802.11 networks using freely accessible software and cheap hardware. Intentional jamming and deauthentication attacks are standalone attacks, but they can be combined with the fake access point attack to increase the latter’s effectiveness. In our research, we work on methods to detect the three different attacks when they are perpetrated independently (one at a time) or concurrently (several at the same time). In this contribution, we present a model that can detect the three attacks, when perpetrated independently, by analysing a set of features (frame interval, Received Signal Strength Indicator, sequence number gap and management frame subtype) extracted from IEEE 802.11 management frame and radiotap headers. We have implemented the model using several supervised learning algorithms. The model with Random Forest and the K-Nearest Neighbour predictors have best detection precision (over 96 %) for fake access point and deauthentication attacks and perfectible detection precision for the intentional jamming attack (over 81%).