arXiv subjects

Michael Beigl

Publications and source records attributed to Michael Beigl.

30 records · Page 2Linked to original sources

arXiv2025

earEOG via Periauricular Electrodes to Facilitate Eye Tracking in a Natural Headphone Form Factor

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 Electrooculography (EOG) eye tracking using 14 electrodes positioned around the ears, integrated into a custom-built headphone form factor device. In a controlled experiment, 16 participants tracked stimuli designed to induce smooth pursuits and saccades. Data analysis identified optimal electrode pairs for vertical and horizontal eye movement tracking, benchmarked against gold-standard EOG and camera-based methods. The electrode montage nearest the eyes yielded the best horizontal results. Horizontal smooth pursuits via earEOG showed high correlation with gold-standard measures ($r_{\mathrm{EOG}} = 0.81, p = 0.01$; $r_{\mathrm{CAM}} = 0.56, p = 0.02$), while vertical pursuits were weakly correlated ($r_{\mathrm{EOG}} = 0.28, p = 0.04$; $r_{\mathrm{CAM}} = 0.35, p = 0.05$). Voltage deflections when performing saccades showed strong correlation in the horizontal direction ($r_{\mathrm{left}} = 0.99, p = 0.0$; $r_{\mathrm{right}} = 0.99, p = 0.0$) but low correlation in the vertical direction ($r_{\mathrm{up}} = 0.6, p = 0.23$; $r_{\mathrm{down}} = 0.19, p = 0.73$). Overall, horizontal earEOG demonstrated strong performance, indicating its potential effectiveness, while vertical earEOG results were poor, suggesting limited feasibility in our current setup.

cs.HC↗

arXiv2025

Heatables: Effects of Infrared-LED-Induced Ear Heating on Thermal Perception, Comfort, and Cognitive Performance

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.

cs.HC↗

arXiv2024

EarCapAuth: Biometric Method for Earables Using Capacitive Sensing Eartips

Earphones can give access to sensitive information via voice assistants which demands security methods that prevent unauthorized use. Therefore, we developed EarCapAuth, an authentication mechanism using 48 capacitive electrodes embedded into the soft silicone eartips of two earables. For evaluation, we gathered capactive ear canal measurements from 20 participants in 20 wearing sessions (12 at rest, 8 while walking). A per user classifier trained for authentication achieves an EER of 7.62% and can be tuned to a FAR (False Acceptance Rate) of 1% at FRR (False Rejection Rate) of 16.14%. For identification, EarCapAuth achieves 89.95%. This outperforms some earable biometric principles from related work. Performance under motion slightly decreased to 9.76% EER for authentication and 86.40% accuracy for identification. Enrollment can be performed rapidly with multiple short earpiece insertions and a biometric decision is made every 0.33s. In the future, EarCapAuth could be integrated into high-resolution brain sensing electrode tips.

cs.CR↗

arXiv2024

OpenEarable ExG: Open-Source Hardware for Ear-Based Biopotential Sensing Applications

While traditional earphones primarily offer private audio spaces, so-called "earables" emerged to offer a variety of sensing capabilities. Pioneering platforms like OpenEarable have introduced novel sensing platforms targeted at the ears, incorporating various sensors. The proximity of the ears to the eyes, brain, and facial muscles has also sparked investigation into sensing biopotentials. However, currently there is no platform available that is targeted at the ears to sense biopotentials. To address this gap, we introduce OpenEarable ExG - an open-source hardware platform designed to measure biopotentials in and around the ears. OpenEarable ExG can be freely configured and has up to 7 sensing channels. We initially validate OpenEarable ExG in a study with a left-right in-ear dual-electrode montage setup with 3 participants. Our results demonstrate the successful detection of smooth pursuit eye movements via Electrooculography (EOG), alpha brain activity via Electroencephalography (EEG), and jaw clenching via Electromyography (EMG). OpenEarable ExG is part of the OpenEarable initiative and is fully open-source under MIT license.

cs.AR↗

arXiv2024

SciEx: Benchmarking Large Language Models on Scientific Exams with Human Expert Grading and Automatic Grading

With the rapid development of Large Language Models (LLMs), it is crucial to have benchmarks which can evaluate the ability of LLMs on different domains. One common use of LLMs is performing tasks on scientific topics, such as writing algorithms, querying databases or giving mathematical proofs. Inspired by the way university students are evaluated on such tasks, in this paper, we propose SciEx - a benchmark consisting of university computer science exam questions, to evaluate LLMs ability on solving scientific tasks. SciEx is (1) multilingual, containing both English and German exams, and (2) multi-modal, containing questions that involve images, and (3) contains various types of freeform questions with different difficulty levels, due to the nature of university exams. We evaluate the performance of various state-of-the-art LLMs on our new benchmark. Since SciEx questions are freeform, it is not straightforward to evaluate LLM performance. Therefore, we provide human expert grading of the LLM outputs on SciEx. We show that the free-form exams in SciEx remain challenging for the current LLMs, where the best LLM only achieves 59.4\% exam grade on average. We also provide detailed comparisons between LLM performance and student performance on SciEx. To enable future evaluation of new LLMs, we propose using LLM-as-a-judge to grade the LLM answers on SciEx. Our experiments show that, although they do not perform perfectly on solving the exams, LLMs are decent as graders, achieving 0.948 Pearson correlation with expert grading.

cs.CL↗

arXiv2024

Explainable Deep Learning Framework for Human Activity Recognition

In the realm of human activity recognition (HAR), the integration of explainable Artificial Intelligence (XAI) emerges as a critical necessity to elucidate the decision-making processes of complex models, fostering transparency and trust. Traditional explanatory methods like Class Activation Mapping (CAM) and attention mechanisms, although effective in highlighting regions vital for decisions in various contexts, prove inadequate for HAR. This inadequacy stems from the inherently abstract nature of HAR data, rendering these explanations obscure. In contrast, state-of-th-art post-hoc interpretation techniques for time series can explain the model from other perspectives. However, this requires extra effort. It usually takes 10 to 20 seconds to generate an explanation. To overcome these challenges, we proposes a novel, model-agnostic framework that enhances both the interpretability and efficacy of HAR models through the strategic use of competitive data augmentation. This innovative approach does not rely on any particular model architecture, thereby broadening its applicability across various HAR models. By implementing competitive data augmentation, our framework provides intuitive and accessible explanations of model decisions, thereby significantly advancing the interpretability of HAR systems without compromising on performance.

cs.AI↗

arXiv2024

MicroNAS: Memory and Latency Constrained Hardware-Aware Neural Architecture Search for Time Series Classification on Microcontrollers

Designing domain specific neural networks is a time-consuming, error-prone, and expensive task. Neural Architecture Search (NAS) exists to simplify domain-specific model development but there is a gap in the literature for time series classification on microcontrollers. Therefore, we adapt the concept of differentiable neural architecture search (DNAS) to solve the time-series classification problem on resource-constrained microcontrollers (MCUs). We introduce MicroNAS, a domain-specific HW-NAS system integration of DNAS, 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 the execution latency and peak memory consumption. Our extensive studies on different MCUs and standard benchmark datasets demonstrate that MicroNAS finds MCU-tailored architectures that achieve performance (F1-score) near to state-of-the-art desktop models. We also show that our approach is superior in adhering to memory and latency constraints compared to domain-independent NAS baselines such as DARTS.

cs.LG↗

arXiv2024

Standardizing Your Training Process for Human Activity Recognition Models: A Comprehensive Review in the Tunable Factors

In recent years, deep learning has emerged as a potent tool across a multitude of domains, leading to a surge in research pertaining to its application in the wearable human activity recognition (WHAR) domain. Despite the rapid development, concerns have been raised about the lack of standardization and consistency in the procedures used for experimental model training, which may affect the reproducibility and reliability of research results. In this paper, we provide an exhaustive review of contemporary deep learning research in the field of WHAR and collate information pertaining to the training procedure employed in various studies. Our findings suggest that a major trend is the lack of detail provided by model training protocols. Besides, to gain a clearer understanding of the impact of missing descriptions, we utilize a control variables approach to assess the impact of key tunable components (e.g., optimization techniques and early stopping criteria) on the inter-subject generalization capabilities of HAR models. With insights from the analyses, we define a novel integrated training procedure tailored to the WHAR model. Empirical results derived using five well-known \ac{whar} benchmark datasets and three classical HAR model architectures demonstrate the effectiveness of our proposed methodology: in particular, there is a significant improvement in macro F1 leave one subject out cross-validation performance.

cs.LG↗

arXiv2023

randomHAR: Improving Ensemble Deep Learners for Human Activity Recognition with Sensor Selection and Reinforcement Learning

Deep learning has proven to be an effective approach in the field of Human activity recognition (HAR), outperforming other architectures that require manual feature engineering. Despite recent advancements, challenges inherent to HAR data, such as noisy data, intra-class variability and inter-class similarity, remain. To address these challenges, we propose an ensemble method, called randomHAR. The general idea behind randomHAR is training a series of deep learning models with the same architecture on randomly selected sensor data from the given dataset. Besides, an agent is trained with the reinforcement learning algorithm to identify the optimal subset of the trained models that are utilized for runtime prediction. In contrast to existing work, this approach optimizes the ensemble process rather than the architecture of the constituent models. To assess the performance of the approach, we compare it against two HAR algorithms, including the current state of the art, on six HAR benchmark datasets. The result of the experiment demonstrates that the proposed approach outperforms the state-of-the-art method, ensembleLSTM.

cs.LG↗

arXiv2022

Universal Distributional Decision-based Black-box Adversarial Attack with Reinforcement Learning

The vulnerability of the high-performance machine learning models implies a security risk in applications with real-world consequences. Research on adversarial attacks is beneficial in guiding the development of machine learning models on the one hand and finding targeted defenses on the other. However, most of the adversarial attacks today leverage the gradient or logit information from the models to generate adversarial perturbation. Works in the more realistic domain: decision-based attacks, which generate adversarial perturbation solely based on observing the output label of the targeted model, are still relatively rare and mostly use gradient-estimation strategies. In this work, we propose a pixel-wise decision-based attack algorithm that finds a distribution of adversarial perturbation through a reinforcement learning algorithm. We call this method Decision-based Black-box Attack with Reinforcement learning (DBAR). Experiments show that the proposed approach outperforms state-of-the-art decision-based attacks with a higher attack success rate and greater transferability.

cs.LG↗

arXiv2022

McXai: Local model-agnostic explanation as two games

To this day, a variety of approaches for providing local interpretability of black-box machine learning models have been introduced. Unfortunately, all of these methods suffer from one or more of the following deficiencies: They are either difficult to understand themselves, they work on a per-feature basis and ignore the dependencies between features and/or they only focus on those features asserting the decision made by the model. To address these points, this work introduces a reinforcement learning-based approach called Monte Carlo tree search for eXplainable Artificial Intelligent (McXai) to explain the decisions of any black-box classification model (classifier). Our method leverages Monte Carlo tree search and models the process of generating explanations as two games. In one game, the reward is maximized by finding feature sets that support the decision of the classifier, while in the second game, finding feature sets leading to alternative decisions maximizes the reward. The result is a human friendly representation as a tree structure, in which each node represents a set of features to be studied with smaller explanations at the top of the tree. Our experiments show, that the features found by our method are more informative with respect to classifications than those found by classical approaches like LIME and SHAP. Furthermore, by also identifying misleading features, our approach is able to guide towards improved robustness of the black-box model in many situations.

cs.LG↗

arXiv2020

Automatic Remaining Useful Life Estimation Framework with Embedded Convolutional LSTM as the Backbone

An essential task in predictive maintenance is the prediction of the Remaining Useful Life (RUL) through the analysis of multivariate time series. Using the sliding window method, Convolutional Neural Network (CNN) and conventional Recurrent Neural Network (RNN) approaches have produced impressive results on this matter, due to their ability to learn optimized features. However, sequence information is only partially modeled by CNN approaches. Due to the flatten mechanism in conventional RNNs, like Long Short Term Memories (LSTM), the temporal information within the window is not fully preserved. To exploit the multi-level temporal information, many approaches are proposed which combine CNN and RNN models. In this work, we propose a new LSTM variant called embedded convolutional LSTM (ECLSTM). In ECLSTM a group of different 1D convolutions is embedded into the LSTM structure. Through this, the temporal information is preserved between and within windows. Since the hyper-parameters of models require careful tuning, we also propose an automated prediction framework based on the Bayesian optimization with hyperband optimizer, which allows for efficient optimization of the network architecture. Finally, we show the superiority of our proposed ECLSTM approach over the state-of-the-art approaches on several widely used benchmark data sets for RUL Estimation.

cs.LG↗