Deep learning has become the dominant paradigm in Wearable Human Activity Recognition (WHAR), yet progress is obscured by a comparability crisis. Results are often reported using inconsistent datasets, custom data processing, and varying evaluation protocols, making state-of-the-art claims fragile. We address this with a large-scale, open-source benchmark that integrates 30 diverse datasets under standardized processing, unified model interfaces, and a shared cross-subject evaluation protocol. Evaluating 17 representative architectures across 4760 training runs, we jointly measure predictive performance alongside on-device latency, peak memory, and model size on an Android reference device. Our results reveal that the WHAR state of the art is distributed rather than dominated by a single architecture. While CNN-HAR achieves the highest mean macro-F1, top-performing models cluster tightly, indicating contemporary architectures have converged near a predictive performance ceiling. When accounting for deployment efficiency, compact neural models, such as TinierHAR, and classical Random Forests define the practically relevant Pareto frontier, whereas larger recurrent and hybrid models incur high hardware costs without corresponding performance gains. Consequently, while predictive performance has plateaued, substantial potential for future progress remains in optimizing deployment efficiency and improving adaptation to domain shifts. We release our full framework to support transparent reuse and extension.

Ear-worn devices are evolving from audio-playback tools into sensing platforms for health, interaction, and context-awareness. Yet, earable systems are typically designed and evaluated under strong assumptions about how long, how often, and in which situations people actually wear personal listening devices (PLDs). To ground these assumptions in-the-wild behavior, we combine a survey of 330 adults with multi-year, passively logged headphone audio-exposure records donated via Apple Health by 90 of them. We characterize where and when people use PLDs, how logged use has changed in recent years, and how psychological traits and social context associate with PLD usage. Our results show that logged mean daily use has increased from 37 minutes in 2020 to 64 minutes in 2024. Listening was intermittent: no listening was logged on 48% of participant-days in 2024, and sessions were fewer but longer on weekends. Sensation seeking, particularly disinhibition, showed small to medium positive associations with self-reported PLD use. Younger adults listened at higher volumes than the 25-34 group. High-volume exposure was uncommon, with only 4% of participant-weeks exceeding World Health Organization (WHO) safe-listening limits. Finally, most participants also reported avoiding PLD use in social situations. We translate these findings into implications for earable computing: realistic expectations of intermittent rather than continuous wear, contextual coverage that anticipates systematic gaps, targeted safe-listening interventions, and personalization grounded in psychosocial and demographic profiles rather than assumptions of uniform use.

Translating natural-language hardware requirements into correct printed circuit board (PCB) schematics remains difficult in embedded, IoT, and wearable development. Designers must choose compatible components, interpret datasheets, add support circuitry, and expose correct interfaces before layout and prototyping can begin, while many such circuits cannot be validated through straightforward simulation. We present pcbGPT, a grounded system for generating editable KiCad schematics from natural-language specifications. pcbGPT represents circuits in a Python DSL and combines tool-augmented synthesis with component-library search, datasheet-grounded design knowledge, execution-based checking, structural and semantic validation, and an interactive web workflow that supports iterative refinement and synchronization with KiCad projects. We evaluate the system on 20 embedded schematic-generation tasks with reference implementations, required components, and interface constraints that enable automatic comparison. The best model reaches overall pass@1 of 0.90 and pass@5 of 1.00; pass@1 is 1.00 on basic and easy tasks, 0.91 on medium tasks, and 0.72 on hard tasks. These results, together with failure analysis, show that pcbGPT can already generate useful, reviewable first-draft schematics for early prototyping, but is not yet reliable enough to replace expert review.

Sensor-based Human Activity Recognition (HAR) models often degrade on unseen users due to domain shifts caused by individual movement patterns and sensor placement. Practical wearable HAR systems therefore require personalization methods that are lightweight, applicable whether calibration data is labeled, unlabeled, or unavailable, and robust under limited calibration. We present a gradient-free framework that repurposes pretrained HAR classifiers as Prototypical Networks using using prior prototypes, which preserve zero-shot performance and regularize adaptation. For labeled calibration, we introduce closed-form Bayesian prototype estimation and extend the same principle to unlabeled calibration. With only 3 seconds of calibration data per activity (one shot), supervised adaptation improves macro-F1 by +2.76 to +33.44 percentage points across four datasets, while unsupervised adaptation improves by +0.56 to +32.13 points. Since adaptation requires only closed-form prototype updates, the framework enables efficient and robust on-device personalization of preexisting HAR classifiers.

Abstract The mosquito Aedes albopictus is expanding across Europe, raising concerns due to its ability to transmit arboviruses. The larvicide Bacillus thuringiensis subsp. israelensis (Bti) plays a central role in controlling the spread of Ae. albopictus. The objective of this study is to assess the temporal effectiveness of Bti as a control method using fine-scale spatiotemporal data and a distributed lag non-linear modelling framework. We analyzed 1,320 ovitrap observations from 195 traps alongside records of 4,387 Bti treatments and local environmental conditions for the entire 2023 mosquito season in Heidelberg, Germany. Bti treatments produced a clear reduction in egg counts, with the strongest marginal effects occurring 6–13 days after treatment with efficacy diminishing thereafter at lower treatment counts, while higher counts demonstrated sustained effects. Cumulatively, Bti was highly effective at reducing egg counts even at moderate treatment levels, and a counterfactual no Bti scenario suggested that Bti treatments reduced seasonal egg production by an estimated 41.9% (95% CI 24.4% – 58.5%) and prevented total establishment across the study site. These results demonstrate that Bti can substantially reduce Ae. albopictus populations in urban settings, though its limited temporal efficacy underscores the need for repeated interventions to prevent establishment and spread.

Chewing side preference (CSP) has been identified both as a risk factor for temporomandibular disorders (TMD) and behavioral manifestation. Despite TMDs affecting roughly one third of the global population, assessment mainly relies on clinical examinations and self-reports, offering limited insight into everyday jaw function. Continuous CSP monitoring could provide an objective proxy for functional asymmetries. Prior wearable approaches, however, mostly use specialized form factors and demonstrate limited performance. We therefore present CHOMP, the first system for chewing side detection using earphones. Employing OpenEarable 2.0, we collected data from 20 participants with microphones, a bone-conduction microphone, IMU, PPG, and a pressure sensor across eleven foods, five non-chewing activities, and three noise conditions. We apply the Continuous Wavelet Transform to each sensing modality and use the resulting multi-channel scalograms as inputs to CNN-based classifiers. Microphones achieve the strongest single-sensor unit performance, with median F1 scores of 94.5

The rise of haptic technology in wearable devices presents new opportunities for affective interventions, promoting mental well-being with minimal effort and immediate relief in everyday situations. This review analyzes 83 relevant articles from 1098 identified, outlining research trends and key dimensions of intervention design and evaluation. Through thematic and descriptive analysis, the studies are organized by affective goals, haptic modalities, body locations, and contextual mechanisms. Three major affective orientations were identified—valence-arousal, social, and embodied affect—revealing a dominant focus on calming and stress-related interventions. Vibrotactile feedback on wearable-friendly areas such as the wrist and forearm was most common, while thermal, pressure, and stroking-based modalities are emerging. Cross-dimensional mapping shows that haptic modality, body placement, and intervention mechanism jointly shape affective outcomes, emphasizing embodied physiological regulation and interpersonal connection as key pathways for affective modulation, as well as future directions involving multi-user interactions. By synthesizing current findings, this work offers a structured framework to guide the design, implementation, and evaluation of affective wearable haptic interventions across diverse application contexts.

Respiratory rate (RR) is a key vital sign for clinical assessment and mental well-being, yet it is rarely monitored in everyday life due to the lack of unobtrusive sensing technologies. In-ear audio sensing is promising due to its high social acceptance and the amplification of physiological sounds caused by the occlusion effect; however, existing approaches often fail under real-world noise or rely on computationally expensive models. We present EarResp-ANS, the first system enabling fully on-device, real-time RR estimation on commercial earphones. The system employs LMS-based adaptive noise suppression (ANS) to attenuate ambient noise while preserving respiration-related acoustic components, without requiring neural networks or audio streaming, thereby explicitly addressing the energy and privacy constraints of wearable devices. We evaluate EarResp-ANS in a study with 18 participants under realistic acoustic conditions, including music, cafeteria noise, and white noise up to 80 dB SPL. EarResp-ANS achieves robust performance with a global MAE of 0.84 CPM , reduced to 0.47 CPM via automatic outlier rejection, while operating with less than 2% processor load directly on the earphone.

Earphones have evolved from pure audio devices to "earables" that are capable of advanced sensing. Bespoke research devices have shown the unique sensing capabilities of the earable platform; however, they are hard to replicate and require expertise to develop in the first place. In this paper, we present OpenEarable 2.0 - an open source, unified platform that integrates a larger number of sensors for conducting comprehensive earable research. OpenEarable 2.0 works as regular binaural Bluetooth earphones and features two ultrasound capable microphones (inward/outward), a 3-axis ear canal accelerometer/bone microphone, a 9-axis head inertial measurement unit, pulse oximeter, optical temperature sensor, ear canal pressure sensor, and microSD card. These capabilities allow for the detection and measurement of 30+ phenomena on the ear that can be used across a wide range of applications in health monitoring, activity tracking, human-computer-interaction and authentication. We describe the design and development of OpenEarable 2.0 which follows best open hardware practices and achieves commercial-level wearability. We provide justification for the selection and placement of integrated sensors and include in-depth descriptions of the extensible, open source firmware and hardware that are implemented using free to use tools and frameworks. For real-time sensor control and data recording we also contribute a web-based dashboard and mobile smartphone app. The wearability and ability to sense different phenomena are validated in four studies which showcases how OpenEarable 2.0 provides accurate measurements in comparison to established gold-standard measurements. We further demonstrate that OpenEarable 2.0 can be assembled by inexperienced users, and that undergraduate students can build applications using the OpenEarable platform.

Eye tracking technology is frequently utilized to diagnose eye and neurological disorders, assess sleep and fatigue, study human visual perception, and enable novel gaze-based interaction methods. However, traditional eye tracking methodologies are constrained by bespoke hardware that is often cumbersome to wear, complex to apply, and demands substantial computational resources. To overcome these limitations, we investigated the application of Electrooculography (EOG) eye tracking using 14 electrodes positioned around the ears, integrated into a custom-built headphone form factor device. In a controlled laboratory experiment, 16 participants tracked a series of on-screen stimuli designed to induce smooth pursuits and saccades. Data analysis identified the optimal electrode pairs for tracking vertical and horizontal eye movements, benchmarked against gold-standard EOG and camera-based eye tracking. The electrode montage closest to the eyes provided the best results for horizontal eye movements. One-dimensional smooth pursuit eye movements measured via earEOG exhibited a high correlation with the gold-standard for horizontal 1D pursuits spanning $$2.5^{\circ }$$ to $$15^{\circ }$$ visual angle for the best performing electrode pair ( $${\textrm{r}}_{\textrm{EOG}} = 0.81, {\textrm{p}}=0.01$$ ; $${\textrm{r}}_{\textrm{CAM}} = 0.56, {\textrm{p}}=0.02$$ ). Vertical 1D smooth pursuits were only weakly correlated for the best performing pair ( $${\textrm{r}}_{\textrm{EOG}} = 0.32, {\textrm{p}}=0.03$$ ; $${\textrm{r}}_{\textrm{CAM}} = 0.31, {\textrm{p}}=0.05$$ ). Voltage deflections of earEOG and gold-standard EOG for saccades from $$2.5^{\circ }$$ to $$15^{\circ }$$ in the four cardinal directions are highly correlated for horizontal eye movement ( $${\textrm{r}}_{\textrm{left}} = 0.99, {\textrm{p}}<0.001$$ ; $${\textrm{r}}_{\textrm{right}} = 1.0, p<0.001$$ ) but not for vertical eye movements ( $${\textrm{r}}_{\textrm{up}} = 0.79, {\textrm{p}}=0.06$$ ; $${\textrm{r}}_{\textrm{down}} = 0.35, {\textrm{p}}=0.53$$ ). A regression model was employed to predict absolute gaze angle changes of horizontal saccades using earEOG and gold-standard EOG. In the left and right directions, the earEOG model achieved a mean absolute angular error of $$3.99 ^\circ \pm 3.45 ^\circ$$ , for saccades ranging from $$2.5^{\circ }$$ to $$15^{\circ }$$ . In comparison, gold-standard EOG attained mean absolute angular error of $$2.98 ^\circ \pm 2.44 ^\circ$$ . Overall, horizontal earEOG demonstrated strong performance, indicating its potential effectiveness in our setup. On the other hand, vertical earEOG showed significantly poorer results, suggesting that it may not be feasible with our current configuration.

The use and research of neural networks on very small processor systems are currently still limited. One of the main reasons is that the design of microcontroller-architecture-aware ML models that take into account user-defined constraints on memory consumption and run-time are very difficult to implement. Therefore, we adapt the concept of differentiable neural architecture search (DNAS) to solve the time series classification problem on resource-constrained microcontrollers (MCUs). This paper explores and demonstrates for the first time that this problem can be solved using Neural Architecture Search (NAS). The key of our specific hardware-aware approach, MicroNAS, is an integration of a DNAS approach, Latency Lookup Tables, Dynamic Convolutions and a novel search space specifically designed for time series classification on MCUs. The resulting system is hardware-aware and can generate neural network architectures that satisfy user-defined limits on execution latency and peak memory consumption. To support our findings, we evaluate MicroNAS under different latency and peak memory constraints. The experiments highlight the ability of MicroNAS to find trade-offs between latency and classification performance across all dataset and microcontroller combinations. As an example, on the UCI-HAR dataset, MicroNAS achieves an accuracy of 94.62% when allowed 25 ms and 98.86% when allowed 50 ms when running on the Nucleo-L552ZE-Q. The much more powerful Arduino Portenta, on the other hand, achieves an accuracy of 95.88% with an allowance of 3 ms and 99.37% when allowed 25 ms displaying the ability of MicroNAS to adapt to different microcontrollers. MicroNAS is also able to find architectures which perform similarly to state-of-the-art systems designed to run on desktop computers (99.62% vs. 99.65% accuracy on the UCI-HAR dataset and 97.83% vs. 97.46% accuracy on the SkodaR dataset).

Automatic sleep staging typically relies on gold-standard EEG setups, which are accurate but obtrusive and impractical for everyday use outside sleep laboratories. This limits applicability in real-world settings, such as home environments, where continuous, long-term monitoring is needed. Detecting sleep onset is particularly relevant, enabling consumer applications (e.g. automatically pausing media playback when the user falls asleep). Recent research has shown correlations between in-ear EEG and full-scalp EEG for various phenomena, suggesting wearable, in-ear devices could allow unobtrusive sleep monitoring. We investigated the feasibility of using single-channel in-ear electrophysiological (ExG) signals for automatic sleep staging in a wearable device by conducting a sleep study with 11 participants (mean age: 24), using a custom earpiece with a dry eartip electrode (Datwyler SoftPulse) as a measurement electrode in one ear and a reference in the other. Ground truth sleep stages were obtained from an Apple Watch Ultra, validated for sleep staging. Our system achieved 90.5% accuracy for binary sleep detection (Awake vs. Asleep) and 65.1% accuracy for fourclass staging (Awake, REM, Core, Deep) using leave-one-subject-out validation. These findings demonstrate the potential of in-ear electrodes as a low-effort, comfortable approach to sleep monitoring, with applications such as stopping podcasts when users fall asleep.

A major challenge of wearable human activity recognition (WHAR) lies in the fact that many existing personalization techniques rely heavily on labeled activity data or physical details of the target subject, which necessitates additional efforts at deployment time. This is often infeasible in practical applications, either due to the lack of input modalities in wearables or due to additional efforts for the users in personalizing their device. To address this problem, we highlight the use of Unsupervised Personalized Deep Learning techniques to enhance the performance of existing Deep Learning WHAR models after initial deployment. This architecture does additional user feedback and can be exploited to improve model performance using only collected sensor data. Our approach identifies samples that present predictive challenges to the original model, trains a surrogate model utilizing data from the training set that is similar to the identified samples, and corrects the predictions of the identified samples through the surrogate model. To validate the effectiveness of the proposed approach, we compare our results against one state-of-the-art Unsupervised Domain Adaptation approach and two state-of-the-art unsupervised personalization approaches using six human activity recognition data sets as benchmarks. Experimental results demonstrate that the proposed method outperforms the state-of-the-art approaches in the unsupervised subject adaptation task. Furthermore, the proposed approach not only attains optimal performance across the majority of data sets but also manifests significant advantages, particularly in terms of stability, across various metrics and data sets.

Maintaining thermal comfort in shared indoor environments remains challenging, as centralized HVAC systems are slow to adapt and standardized to group norms. Cold exposure not only reduces subjective comfort but can impair cognitive performance, particularly under moderate to severe cold stress. Personal Comfort Systems (PCS) have shown promise by providing localized heating, yet many designs target distal body parts with low thermosensitivity and often lack portability. In this work, we investigate whether targeted thermal stimulation using in-ear worn devices can manipulate thermal perception and enhance thermal comfort. We present Heatables, a novel in-ear wearable that emits Near-Infrared (NIR) and Infrared (IR) radiation via integrated LEDs to deliver localized optical heating. This approach leverages NIR-IR's ability to penetrate deeper tissues, offering advantages over traditional resistive heating limited to surface warming. In a placebo-controlled study with 24 participants, each exposed for 150 minutes in a cool office environment (approximately 17.5 degrees Celsius) to simulate sustained cold stress during typical sedentary office activities, Heatables significantly increased the perceived ambient temperature by around 1.5 degrees Celsius and delayed cold discomfort. Importantly, thermal benefits extended beyond the ear region, improving both whole-body comfort and thermal acceptability. These findings position in-ear NIR-IR-LED-based stimulation as a promising modality for unobtrusive thermal comfort enhancement in everyday contexts.

The lack of standardization across Wearable Human Activity Recognition (WHAR) datasets limits reproducibility, comparability, and research efficiency. We introduce WHAR datasets, an open-source library designed to simplify WHAR data handling through a standardized data format and a configuration-driven design, enabling reproducible and computationally efficient workflows with minimal manual intervention. The library currently supports 9 widely-used datasets, integrates with PyTorch and TensorFlow, and is easily extensible to new datasets. To demonstrate its utility, we trained two state-of-the-art models, TinyHar and MLP-HAR, on the included datasets, approximately reproducing published results and validating the library's effectiveness for experimentation and benchmarking. Additionally, we evaluated preprocessing performance and observed speedups of up to 3.8x using multiprocessing. We hope this library contributes to more efficient, reproducible, and comparable WHAR research.

In this demo, we present OpenEarable 2.0, an open-source earphone platform designed to provide an interactive exploration of physiological ear sensing and the development of AI applications. Attendees will have the opportunity to explore real-time sensor data and understand the capabilities of OpenEarable 2.0's sensing components. OpenEarable 2.0 integrates a rich set of sensors, including two ultrasound-capable microphones (inward/outward), a 3-axis ear canal accelerometer/bone conduction microphone, a 9-axis head inertial measurement unit, a pulse oximeter, an optical temperature sensor, an ear canal pressure sensor, a microSD slot, and a microcontroller. Participants will be able to try out the web-based dashboard and mobile app for real-time control and data visualization. Furthermore, the demo will show different applications and real-time data based on OpenEarable 2.0 across physiological sensing and health monitoring, movement and activity tracking, and human-computer interaction.

Human activity recognition (HAR) research often lacks accessible, comprehensive field data. Commercial systems are rarely open source, hard to expand, and limited by issues like node synchronisation, data throughput, unclear sensor placement, complexity, and high cost. As a result, researchers typically use only a few intuitively placed sensors and conduct limited field trials. HARNode overcomes these challenges with a fully open-source hardware and software platform. Each node includes an ESP32-S3 module (AtomS3), a 9-axis IMU (Bosch BMX160), pressure and temperature sensors (Bosch BMP388), a display, and an I2C port. Data is streamed via Wi-Fi, with NTP-based time synchronisation achieving 1 ms accuracy. The system runs for up to 8 hours and is built using off-the-shelf parts, a simple online PCB service, and a compact 3D-printed housing with Velcro straps, enabling flexible and scalable body placement while requiring little hardware knowledge. In a study with ten subjects wearing eleven HARNodes each, setup took under five minutes per person. A random forest classifier distinguished walking from stair-climbing transitions, showing the benefits of sensor-overprovisioning: Seven nodes achieved approximate to 98% accuracy, matching the performance of all eleven. These findings confirm HARNode's value as a fast-deploying, scalable tool for field-based HAR research and optimised sensor placement.

The demand for next-generation flexible electronics in applications like smart packaging and smart bandages has driven the need for cost-effective solutions. Traditional silicon-based electronics struggle with high costs and rigidity, making them unsuitable for these emerging markets. In this regard, additive printed electronics (PE) offer a viable alternative with their flexibility and ultra-low-cost manufacturing. printed analog neuromorphic circuits (pNCs) are well-suited for these target applications, especially for classification tasks, as their low device count can efficiently meet the needs of the technology. However, low-cost additive manufacturing comes with higher defect rates, such as misprints, broken connections, and defective components, posing significant challenges to the reliability of printed circuits. This article presents a novel co-design of training algorithm and hardware for fault-tolerant pNCs using fault-aware training (FAT). The proposed method introduces a fault-tolerant version of printed nonlinear transformation circuits, combined with a bespoke training process that selects different types of printed activation functions (AFs) for different neurons to optimize both fault endurance and hardware costs. Experiments on benchmark datasets demonstrate an improvement in the accuracy of fault-tolerant (FT) pNCs from 62.1% to 79.4% under a 10% fault rate. Moreover, combining both normal and fault-tolerant versions of activation functions (AFs) using gumble-softmax distribution shows an acceptable accuracy drop with an average reduction in power and area of 54.5% and 6.54%, respectively, while reducing the training time significantly by 56.2%, compared to only using FT-AFs.

Interaction with earables - earphones equipped with additional sensors - has been identified as one of four major areas of earable research. Worn naturally and positioned near key physiological signals, earables support a wide range of interaction modalities and have demonstrated the ability to detect multiple inputs simultaneously. Yet this diversity has resulted in a fragmented body of research, making it increasingly difficult to track developments and identify relevant studies. To address this, we introduce EarXplore, a curated, interactive online database on earable interaction research. Designed through a question-centered process that guided both the development of 34 criteria applied to annotate 118 studies and the structure of the platform, EarXplore comprises four distinct yet integrated views: a Tabular View for structured exploration, a Graphical View for visual overviews, a Similarity View for identifying conceptual links, and a Timeline View for analyzing trends and scholarly lineage. We demonstrate how the platform supports tailored exploration, targeted filtering, and interactive information retrieval, allowing researchers to query the literature and synthesize information in the format of their choice. We furthermore leverage the contents and capabilities of the platform to discuss the research gaps and opportunities in the field. With built-in mechanisms for continuous community updates, EarXplore not only reflects the current state of the field but also evolves alongside it, serving as a living resource to inform and accelerate future developments.

Wearable haptic interventions can support relaxation through slow, vibrotactile biofeedback. However, most applications focus on stress-inducing tasks and fixed vibration patterns, with limited attention to body placement or dynamic feedback during restful states. This study examined real-time heart rate-driven vibrotactile feedback during eyes-closed wakeful rest at four body locations: wrist, hand, forearm, and shoulder. We measured heart rate, ear-based alpha EEG, subjective restfulness, and user experience. Biofeedback reduced heart rate at the wrist, shoulder, and forearm, while alpha power remained unchanged. Restfulness ratings and location preference were highest at the forearm and shoulder. Participants also reported greater comfort, relaxation, and sleepiness at the forearm, while the wrist was more noticeable. These findings suggest the forearm and shoulder are well-suited for unobtrusive relaxation feedback, whereas the wrist may benefit from design refinement.