BIOSIGNALS 2025 Abstracts

Full Papers Short Papers
Full Papers
Paper Nr: 24
Title:

A Novel Cuff-Less and Calibration-Free Blood Pressure Estimation Framework Using Single Photoplethysmogram
Authors:

Yusuf Ziya Hayirlioglu and Beren Semiz
Abstract: Blood pressure (BP) is one of the four main vital signs and is a key indicator of cardiovascular health. However, monitoring of BP is not regularly done in most of the population until health problems arise. Continuous and convenient monitoring of blood pressure is thus needed to address this issue. We propose a novel BP estimation algorithm without calibration to estimate BP from a cuff-less photoplethysmogram (PPG) system. Data from a total of 219 subjects, which underwent only simple preprocessing steps, was used to train and evaluate a hybrid Convolutional Long Short-Term Memory Neural Network (CNN-LSTM) model. The model was trained using the preprocessed PPG signal as the only input. The model had two neurons in the last layer to output systolic blood pressure (SBP) and diastolic blood pressure (DBP) values. The model was optimized by conducting a random search on its hyperparameters for better performance. The model resulted in a comparable performance to those in the literature, with mean absolute errors (MAEs) of 14.13 mmHg and 8.80 mmHg for SBP and DBP, respectively. To assess generalizability, we also tested the trained model on a second dataset collected from 20 subjects using a custom wearable system, which was again resulted in MAEs of 10.71 mmHg and 10.09 mmHg, respectively. Overall, our results show that such a pipeline could potentially be leveraged in the design of wearable systems to achieve cuff-less and calibration-free BP monitoring in ambulatory settings.
Download
Paper Nr: 28
Title:

Comparison Between Machine Learning and Deep Learning on Multiple Motor Imagery Paradigms in a Low-Resource Context
Authors:

Langlois Quentin and Jodogne Sébastien
Abstract: Motor Imagery (MI) decoding is a task aimed at interpreting the mental imagination of movement without any physical action. MI decoding is typically performed through automated analysis of electroencephalographic (EEG) signals, which capture electrical activity of the brain via electrodes placed on the scalp. MI decoding holds significant potential for controlling devices or assisting in patient rehabilitation. In recent years, Deep Learning (DL) techniques have been extensively studied in the MI decoding domain, often outperforming traditional Machine Learning (ML) methods. However, these DL models are known to require large amounts of data to achieve good results and substantial computational resources, limiting their applicability in low-data or low-resource contexts. This work explores these assumptions by comparing state-of-the-art ML and DL models under simulated low-resource conditions. Experiments were conducted on the Kaya2018 dataset, enabling this comparison across multiple MI paradigms, which contrasts with other studies that typically focus only on left/right-hand decoding task. The results indicate that even with limited data, DL models consistently outperform ML techniques across all evaluated MI tasks, with the most significant advantage observed in advanced experimental setups.
Download
Paper Nr: 116
Title:

Advancements in Wearable EEG Technology: Electrode Characterization and Signal Quality Assessment
Authors:

Andrea Farabbi, Andrea Costanzo Palmisciano, Matteo Rossi, Niccolò Antonello, Diana Trojaniello, Tommaso Ongarello, Pietro Cerveri and Luca Mainardi
Abstract: This research contributes to the advancement of practical, user-friendly EEG devices for both research and real-world applications. The paper presents a comprehensive study on the development and characterization of wearable electroencephalography (EEG) recording in non-traditional electrode locations. In particular, we focus on optimizing electrode placement, material selection, and signal quality assessment. Our investigation includes impedance testing of various electrode materials, comparative analysis of dry versus wet electrodes, and validation through standard EEG protocols. Results demonstrate the feasibility of acquiring high-quality EEG signals from over-the-ear locations where using gold-plated brush electrodes with retractile pins, show superior impedance characteristics (105Ω) compared to other tested materials. We also validate and compare dry electrodes by means of an eyes-open/eyes-closed protocol, confirming the ability to detect alpha rhythm modulation in non-traditional electrode placements.
Download
Paper Nr: 124
Title:

Transfer-Modal Extraction of Surface EMG Features for Upper Limb Motor Classification
Authors:

Vedant Mangrulkar and Madhav Rao
Abstract: Surface Electromyography (sEMG) signals provide critical insights into muscular activity, aiding action classification and monitoring muscular disorders. However, their reliability is hindered by noise and unstructured data. Despite the advancements in machine learning, large datasets are essential to address these challenges and enhance decoding accuracy for further development. Hence, this work attempts to predict the sEMG features from the accelerometer signals in a view to generate synthetic data which is useful for further developments around this physiological signal. This work examines the correlation between accelerometer-generated sEMG features and those from original sEMG signals for four upper limb actions wrist flexion, wrist extension, wrist closing and wrist vibration focusing on the flexor carpi ulnaris and extensor carpi radialis muscles. Synthesized features are augmented with original features to train an ML model, achieving 91% accuracy on unseen original sEMG features. This work showcases a viable solution to generate more sEMG features corresponding to the actions under test from an altogether different modality. This work is a step towards synthesizing EMG signals and features for human limb movements which offers a strong platform to design imitation learning for rehabilitation systems in the future.
Download
Paper Nr: 138
Title:

Analyzing Cognitive Patterns in Gifted Children Using MRI and Morphometric Similarity Networks
Authors:

Shuning Han, Feng Duan, Gemma Vilaseca, Núria Vilaró, Cesar F. Caiafa, Zhe Sun and Jordi Solé-Casals
Abstract: Advances in non-invasive neuroimaging, such as structural magnetic resonance imaging (sMRI), have enabled the construction of structural brain networks (SBNs), allowing in vivo mapping of anatomical connections. This study investigates brain network structural differences linked to different intelligence levels in children by individual morphometric similarity networks (MSNs) derived from sMRI data. Through group- and individual-level analyses, we aim to uncover key topological features associated with cognitive performance and to identify a suitable connection density for SBN analysis. Connection density strongly affects global and nodal topological features, with a range of p = 0.05 to 0.15 recommended for stable and optimal results. Gifted individuals exhibit stronger intra-hemispheric and intra-modular connectivity, a more balanced distribution of left-to-right intra-hemispheric connections, and lower mean versatility, supporting efficient and stable cognitive processing. Moreover, anatomical modularity analyses based on von Economo indicate that higher cognitive performance is linked to enhanced connectivity in specific areas (such as secondary sensory area, motor to association area and secondary sensory to limbic area), alongside selective reduction in certain modular connections (such as motor to insular area, association to secondary sensory area and motor to secondary sensory area). Furthermore, topological features, including participation coefficient and local efficiency, are linked to cognitive performance. These findings provide valuable insights into the SBNs underlying cognitive levels in children.
Download
Paper Nr: 140
Title:

Multi-Scale Probabilistic Score Fusion for Enhancing Alzheimer’s Disease Detection Using EEG
Authors:

Maxime Bedoin, Bernadette Dorizzi, Jérôme Boudy, Kiyoka Kinugawa and Nesma Houmani
Abstract: In this work, we propose a fusion approach to analyze EEG signals for the discrimination between patients with Subjective Cognitive Impairment (SCI) and patients suffering from Alzheimer’s Disease (AD). In this framework, we analyze EEG signals at different epoch durations, following a multi-scale procedure, and in different frequency bands, using Phase-Lag Index (PLI) and Dynamic Time Warping (DTW) for functional connectivity measurement. Experiments show that our fusion approach leads to an improvement of classification results, reaching an AUC of 0.902 with PLI, and 0.894 with DTW; whereas we obtain an AUC of 0.845 with PLI and 0.801 with DTW when computing connectivity on the entire signal, as usually done in the literature. Furthermore, with the additional fusion of the scores obtained at different frequency bands, we reach the best performance with both PLI (AUC=0.95, Accuracy=91%) and DTW (AUC=0.98, Accuracy=95%). Finally, we investigate the generalization capability of our proposal on a test set. We found that our fusion scheme allows obtaining better classification results comparatively to when we consider the entire signal to compute functional connectivity.
Download
Paper Nr: 199
Title:

Modulating Cerebral Rhythms in Parkinson's Disease: Insights on the Role of Auditory Stimulation
Authors:

Pablo García-Peña, Juan M. López, Milagros Ramos, Daniel González-Nieto and Guillermo de Arcas
Abstract: Parkinson's disease (PD) is a neurodegenerative disease characterized by a slowing of brain rhythms, leading to cognitive and motor deficits. Auditory stimulation (AS) has proven to have great potential as a non-invasive approach to modulate brain activity. Nevertheless, despite promising clinical and preclinical findings, optimal AS parameters for its use in PD remain unclear. To investigate the potential therapeutic effects of AS in PD, we aimed to establish an optimal preclinical model and stimulation protocol. Two mouse strains were compared, CD1 and C57BL/6, and assessed their auditory sensitivity. 3-months C57BL/6 mice was selected as the most suitable model for auditory studies. Two literature-based AS protocols were applied, a 10 kHz carrier tone modulated with 40 Hz pulses and a 40 Hz amplitude-modulated tone. Our results demonstrate that comparing pre- and post-stimulation periods, the 10 kHz/40 Hz protocol consistently induced a reduction in delta power and an increase in gamma relative power, with persistent effects of the latter 24 hours post-stimulation. These findings suggest that this specific AS protocol holds promise for targeting abnormal brain rhythms associated with PD and may have potential therapeutic implications. Further research is needed to explore the underlying mechanisms and optimize AS parameters for clinical translation.
Download
Paper Nr: 215
Title:

Novel and Efficient Hyperdimensional Encoding of Surface Electromyography Signals for Hand Gesture Recognition
Authors:

Ancelin Salerno and Sylvain Barraud
Abstract: Gesture recognition has become a crucial component of human-computer interaction, with applications ranging from virtual reality to assistive technologies. This study explores Hyperdimensional Computing (HDC) as a powerful alternative to traditional machine learning techniques for real-time gesture recognition. HDC is known for its robustness and efficiency, enabling fast and accurate classification though the use of high-dimensional binary vectors. In this study, we introduce two key variants aimed at significantly improving the performance of gesture recognition: (1) an enhancement of item memory representation enabling a better gestures recognition, and (2) an advanced temporal encoding mechanism that captures the dynamic nature of gestures more efficiently. These modifications are evaluated using a benchmark dataset of surface electromyography (sEMG) signals, demonstrating significant improvements in both accuracy and computational efficiency.
Download
Paper Nr: 219
Title:

Automatic Ocular Artifact Correction in Electroencephalography for Neurofeedback
Authors:

Cassandra Dumas, Marie Constance Corsi, Claire Dussard, Fanny Grosselin and Nathalie George
Abstract: Ocular artifacts can significantly impact electroencephalography (EEG) signals, potentially compromising the performance of neurofeedback (NF) and brain-computer interfaces (BCI) based on EEG. This study investigates if the Approximate Joint Diagonalization of Fourier Cospectra (AJDC) method can effectively correct blink-related artifacts and preserve relevant neurophysiological signatures in a pseudo-online context. AJDC is a frequency-domain Blind Source Separation (BSS) technique, which uses cospectral analysis to isolate and attenuate blink artifacts. Using EEG data from 21 participants recorded during a NF motor imagery (MI) task, we compared AJDC with Independent Component Analysis (ICA), a widely used method for EEG denoising. We assessed the quality of blink artifact correction, the preservation of MI-related EEG signatures, and the influence of AJDC correction on the NF performance indicator. We show that AJDC effectively attenuates blink artifacts without distorting MI-related beta band signatures and with preservation of NF performance. AJDC was calibrated once on initial EEG data. We therefore assessed AJDC correction quality over time, showing some decrease. This suggests that periodic recalibration may benefit long EEG recording. This study highlights AJDC as a promising real-time solution for artifact management in NF, with the potential to provide consistent EEG quality and to enhance NF reliability.
Download
Paper Nr: 224
Title:

Short-Term Effects of Mindful Uni-Nostril Breathing on Cardio-Autonomic Functions: A Randomized Controlled Trial
Authors:

Satyam Tiwari and Arnav Bhavsar
Abstract: Uni-nostril mindful breathing, an ancient yogic practice, has been suggested to influence autonomic nervous system function differentially, yet systematic evidence remains limited. This randomized controlled trial investigated the effects of nostril-specific breathing techniques on autonomic nervous system modulation in healthy adults. Ninety participants were randomly assigned to one of three groups: left-nostril breathing, right-nostril breathing, or a control group performing unstructured breathing for 10 minutes. HRV parameters and systolic and diastolic blood pressures were collected pre-and post-intervention. Left nostril breathing significantly decreased HRV parameters (SDNN: -27.0%, RMSSD: -25.1%) while increasing SI (+37.4%) and SNS activity (+98.7%), therefore suggesting increased sympathetic activation. With little impact on other autonomic indicators, right-nostril breathing showed significant decreases in both systolic (-5.5 mmHg) and diastolic blood pressure (-3.3 mmHg). These results support nostril-specific breathing as a simple, non-pharmacological technique for autonomic modulation, offering prospective applications in stress and cardiovascular management, with varying effects dependent upon nostril selection.
Download
Paper Nr: 256
Title:

Assessing Electrocardiogram Quality: A Deep Learning Framework For Noise Detection And Classification
Authors:

Márcia Monteiro, Mariana Dias and Hugo Gamboa
Abstract: The electrocardiogram (ECG) is an essential tool in the diagnosis of cardiovascular conditions. A common obstacle to readability and reliability is the vulnerability of ECG signals to noise, especially in wearable devices and long-term monitoring. Traditional methods have limited accuracy in noise detection, and, while deep learning (DL) shows promise, current models primarily focus on binary classification, lacking detailed quality analysis. This study proposes a DL model that assesses ECG signal quality, detecting and classifying specific noise types, with random-length noise segments added to clean 10-second signals to simulate real-world scenarios. The model, using gated recurrent units (GRUs), identifies three common noise types: baseline wander (BW), muscle artifacts (MA), and electrode motion (EM), achieving 98.09 % accuracy for BW, 92.62 % for MA, and 90.71 % for EM with F1 scores of 88.89 % for BW, 82.19 % for EM and 64.62 % for MA. It also surpasses existing DL methods, reaching 99.86 % accuracy for binary classification, with high recall and precision.
Download
Paper Nr: 287
Title:

Investigation of the Relational Strength Between Suspected Atrial Fibrillation Triggers and Detector-Based Arrhythmia Episode Occurrence
Authors:

Vilma Pluščiauskaitė and Andrius Petrėnas
Abstract: Atrial fibrillation (AF) treatment remains challenging, with current options often limited to anticoagulants and antiarrhythmic medications. Growing evidence suggests that acute exposures, referred to as AF triggers, can initiate AF in some patients. Therefore, identifying and managing personal triggers may serve as an effective strategy to complement conventional treatment. This study explores the utility of wearable-based biosignals to assess the relational strength between the suspected triggers and AF occurrence when episodes are detected using electrocardiogram (ECG) and photoplethysmogram (PPG). Biosignals from 33 patients with paroxysmal AF (mean age 61 ± 13 years), who wore an ECG patch and a wrist-worn PPG device during a 7.0 ± 0.7 day observation period, were used in the study. Suspected triggers due to physical exertion, psychophysiological stress, and lying on the left side were identified based on a detection parameter calculated over successive segments of the ECG and/or acceleration signals. The relational strength between a suspected trigger and AF episodes is quantified based on AF burden, defined as the ratio of time spent in AF to the total analysis time interval, assuming that the post-trigger AF burden is greater than the pre-trigger AF burden. The results indicate that the relational strength between suspected triggers and AF episode occurrence, as detected using ECG- and PPG-based AF detectors, differs from manual annotation by an average of 0.03±0.15 and -0.21±0.21, respectively. This study demonstrates the potential of wearable-based biosignals in providing personalized identification of suspected AF triggers. However, challenges such as non-wear periods and poor PPG signal quality remain to be addressed for practical applications.
Download
Paper Nr: 324
Title:

Histopathological Imaging Dataset for Oral Cancer Analysis: A Study with a Data Leakage Warning
Authors:

Marcelo Nogueira and Elsa Ferreira Gomes
Abstract: Oral squamous cell carcinoma is one of the most prevalent and lethal types of cancer, accounting for approximately 95% of oral cancer cases. Early diagnosis increases patient survival rates and has traditionally been performed through the analysis of histopathological images by healthcare professionals. Given the importance of this topic, there is an extensive body of literature on it. However, during our bibliographic research, we identified clear cases of data leakage related to contamination of test data due to the improper use of data augmentation techniques. This impacts the published results and explains the high accuracy values reported in some studies. In this paper, we evaluate several models, with a particular focus on EfficientNetBx architectures combined with Transformer layers, which were trained using Transfer Learning and Data Augmentation to enhance the model’s feature extraction and attention capabilities. The best result, obtained with the Effi-cientNetB0, together with the Transformer layers, achieved an accuracy rate of 87.1% on the test set. To ensure a fair comparison of results, we selected studies that we identified as not having committed data leakage.
Download
Paper Nr: 337
Title:

Greedy Brain Source Localization with Rank Constraints
Authors:

Viviana Hernandez-Castañon, Steven Le Cam and Radu Ranta
Abstract: This paper introduces a new low rank matrix approximation model and a greedy algorithm from the iterative regression family to solve it. Unlike the classical Orthogonal Matching Pursuit (OMP) or Orthogonal Least Squares (OLS), the elements of the dictionary are not vectors but matrices. For reconstructing a measurement matrix from this dictionary, the regression coefficients are thus matrices, constrained to be low (unit) rank. The target application is the inverse problem in brain source estimation. On simulated data, the proposed algorithm shows better performances than classical solutions used for solving the mentioned inverse problem.
Download
Short Papers
Paper Nr: 33
Title:

Electroencephalography Analysis Frameworks for the Driver Fatigue Problem: A Benchmarking Study
Authors:

Kemalcan Kucuk, Efe Ismet Yurteri and Beren Semiz
Abstract: Driver fatigue problem is a major factor contributing to traffic accidents globally, making its analysis and detection crucial for early prevention. Among various approaches for detecting driver fatigue, electroencephalography (EEG) processing is one of the most widely employed techniques. This study investigates different feature extraction and machine learning methodologies for detecting driver fatigue using EEG signals, and provides a comparative performance analysis against existing methods. To that aim, we used a publicly available dataset collected during a simulated driving task and applied our feature extraction methods to the concurrently recorded EEG signals. Various features from distinct groups were extracted to serve as the foundation for subsequent analyses. The 30 channels from the original dataset were individually evaluated based on the performance of machine learning algorithms trained on each channel, allowing for the selection of the four most optimal channels. Using these selected channels, the different subsets of extracted features were then compared based on their accuracy values. For further analysis, the features were ranked using both ANOVA and Chi-Squared feature selection methods to examine the impact of the number of features on model performance. Each model was first trained using a standard training-testing split, where the highest-scoring model was a Support Vector Machine (SVM) achieving a test accuracy of 90.73%. Additionally, using a Leave-One-Out Cross-Validation (LOOCV) approach, the highest performing model was found to be the k-Nearest Neighbors (K-NN) classifier with an average test accuracy of 70.45%. The analyses and comparisons presented in this study may serve as a basis for developing real-time applications and for further in-depth investigations.
Download
Paper Nr: 35
Title:

Preliminary Technical Test of Different Physiological Modalities to Detect Workload in Humans in Microgravity
Authors:

Judith Bütefür and Elsa Andrea Kirchner
Abstract: In this work we aim to investigate whether eye tracking, electrocardiogram and respiration are good measures to detect workload (WL) of humans in microgravity. To this end, an auditory N-back study was performed during a parabolic flight in microgravity and during a control condition in the lab under Earth gravity by 3 operators of the experiment. The data were analysed regarding their predictive nature to estimate WL. The results show that none of the parameters are suitable for WL detection in humans due to the very short microgravity phases (~22s) and due to scopolamine intake. Nevertheless, some parameters are potentially suitable for longer stay in microgravity. In addition, the results of this study were compared with the results of a previously published electroencephalogram (EEG) analysis on the same data set. This comparison shows that EEG is a more promising predictor modality for WL. In future work, we will conduct another study to extend the number of operators. Different conditions than short term parabolic flights and measurement with longer duration are needed to investigate the stability of WL prediction.
Download
Paper Nr: 67
Title:

Quantifying Racial Bias in SpO 2 Measurements Using a Machine Learning Approach
Authors:

Hakan Burak Karli, Eli Hilborn and Bige Deniz Unluturk
Abstract: This paper investigates the racial biases in pulse oximetry, focusing on the importance of noninvasive peripheral oxygen saturation (SpO2) measurements in classifying patient race and ethnicity. Using the publicly available BOLD dataset, our study applies various machine learning models to quantify the extent of bias in SpO2 readings. Initial analysis revealed significant inaccuracies for individuals with darker skin tones, highlighting broader health disparities. Further exploration with machine learning models assessed SpO2 as a predictive marker for race, uncovering that conventional oximetry may underestimate hypoxemia in non-White patients. Notably, the XGBoost model demonstrated superior performance, achieving baseline accuracy of 58.08% across the dataset with all races and 72.60% for only black and white patients included, while consis-tently identifying SpO2 as a significant factor in these disparities. Our findings demonstrate the necessity for recalibrating medical devices to enhance their reliability and inclusivity, ensuring equitable health outcomes.
Download
Paper Nr: 90
Title:

Unveiling Vocal Phenotypes of Dysphonia with Unsupervised Learning
Authors:

Federico Calà, Francesco Correnti, Lorenzo Frassineti, Giovanna Cantarella, Giulia Buccichini, Ludovica Battilocchi and Antonio Lanatà
Abstract: Dysphonia is a voice disorder caused by morphological and neurological alterations. This work proposes a clustering analysis on vocal properties of patients diagnosed with benign lesions of the vocal folds (BLVF) and unilateral vocal fold paralysis (UVFP) to identify if they constitute separate vocal subtypes of dysphonia and to understand whether misclustered data depend on a specific diagnosis and age. Two hundred seventy-five patients uttered a sustained vowel /a/, from which acoustic features were extracted and transformed. Two conditions were tested separately for each gender: the unaware and the aware approach, where statistical analysis was performed to select the significantly different parameters between BLVF and UVFP. The best clustering results were obtained for the aware condition, with a silhouette score of 0.70 for both genders; accuracies were 0.67 and 0.70 for the female and male patients. A single component was retained for both genders: phonation and articulation parameters presented high weights for female and male patients, respectively. Misclustered observations analysis showed that feature transformation and reduction improved the UVFP voices clusterability. The clustering error outcome did not depend on age, voice disorder types, or subtypes. These findings may contribute to a better understanding of voice disorders’ properties, reducing misdiagnoses and supporting their follow-up.
Download
Paper Nr: 120
Title:

Automatic Classification of Parkinson’s Disease Through the Fusion of Sustained Vowel Descriptors
Authors:

Sahar Hafsi, Linda Marrakchi-Kacem, Farouk Mhamdi and Sonia Djaziri-Larbi
Abstract: Voice disorders are early symptoms of Parkinson’s disease (PD) and have motivated the use of speech as a biomarker for PD. In particular, dysfunctional phonation of sustained vowels has gained increasing interest in the automatic classification of PD. However, most studies typically focus on a single vowel to extract disease descriptors, which may limit the detection of subtle vocal alterations present in PD patients. The main objective of this study is to investigate the contribution of analyzing two vowels for the automatic classification of PD, as opposed to relying on a single vowel. In this paper, we propose a novel automatic approach to identify dysphonia in PD by combining speech descriptors extracted from two sustained vowels, /a:/ and /i:/. This fusion enables the detection of a broader range of vocal alterations, thereby increasing the robustness of the predictive models. A preprocessing of the speech signals was performed, followed by feature selection using the ReliefF algorithm. Then, a robust nested cross-validation was applied to evaluate the models. The results clearly indicate higher classification performance when combining the descriptors of /a:/ and /i:/.
Download
Paper Nr: 132
Title:

Comparative Analysis of Generalized Multiscale Entropy Methods for Coarse-Grained Time Series Construction in Assessing Autonomic Balance in Peripheral Arterial Disease Patients
Authors:

O. Barquero-Pérez, R. Goya-Esteban, E. Sarabia-Cachadiña and J. Naranjo-Orellana
Abstract: Peripheral Arterial Disease (PAD) is a chronic condition that significantly impacts autonomic balance, as reflected in Heart Rate Variability (HRV). However, the characterization of autonomic balance in PAD patients using HRV is still unclear. Generalized Multiscale Entropy (GMSE) is a nonlinear method capable of characterizing the complexity of HRV across multiple time scales, offering a more nuanced understanding of autonomic dysfunction in PAD patients. 14 healthy male subjects (60±5 years) and 14 male intermittent claudication patients (64±6 years) underwent 10 minutes of ECG recording from which RR interval time series were obtained. This study provides a comparative analysis of different GMSE methods for constructing coarse-grained time series, specifically using the mean, mean absolute deviation (MAD), standard deviation (σ), and variance (σ2) approaches. By applying these methods, we investigate their efficacy in differentiating between healthy individuals and PAD patients. Our results demonstrate that the variance coarse-grained method offers superior discriminatory power, revealing statistically significant differences. These findings suggest that the variance-based GMSE method is the most effective approach for assessing autonomic imbalance in PAD patients, with potential applications in improving diagnostic tools and treatment strategies.
Download
Paper Nr: 148
Title:

Impact of Feature Extraction Optimization on Machine Learning Models for sEMG-Based Prosthesis Control
Authors:

Ricardo Henrique Avelar Matheus and Maria Claudia F. Castro
Abstract: One of the most significant challenges to the quality of life for amputees is the development of prostheses that can closely simulate the capabilities of the lost limb. One possible solution to this problem is myoelectric prostheses, which are devices that use myolectric signals as users’ intention to perform independent movements. This study aims to investigate how optimizing feature extraction methods can impact the performance of machine learning models in recognizing surface electromyogram (sEMG) signals from amputees. The LibEMG library in Python, which offers a simple and robust API for developing sEMG-based projects, was used alongside the DB8 dataset from the NINAPRO public database, which promotes machine-learning research in human, robotic, and prosthetic hands. A total of twelve feature extraction methods and seven different classifiers were tested. The results showed the best mean accuracy of 79.18% using a Random Forest classifier with a set of eleven time and frequency domain features, considering the data of an amputee with experience in using myoelectric prostheses. However, the most affected models by feature optimization were KNN, MLP, and SVM, with accuracy improvements up to 69.28%.
Download
Paper Nr: 149
Title:

Exploring the Relationship Between Intracavitary Electrohysterogram Characteristics from Contraction and Window Analysis
Authors:

Juan Miguel Mira-Tomas, Alba Diaz-Martinez, Jose Alberola-Rubio, Pilar Alamá Faubel, Gemma Castillón Cortés, Sergio Caballero Sanz and Javier Garcia-Casado
Abstract: Assisted reproductive technologies are increasingly common due to the rising maternal age. One potential cause of embryo implantation failure is altered uterine peristalsis patterns. Intracavitary electrohysterography (IC-EHG) is a recent technique developed to characterize the electrophysiology of uterine peristalsis throughout the menstrual cycle. Two primary methodologies are employed for analysis: Contraction Analysis and Window Analysis. This study aims to examine the relationship between parameters describing the same characteristics of the signals using contraction and window analysis of 2, 4 and 10 minutes. Peristalsis was recorded at three different menstrual cycle phases from 10 fertile healthy women. Continuous 10 minutes recordings free of artifacts were selected. A very strong linear relationship (R2 ≥ 0.95) was found between the amplitude parameter from contraction (Root Mean Square (RMS)) and window (80th percentile of signal RMS envelope) analysis. For the spectral parameter (Median Frequency), the relationship was strong (0.59 ≤ R2 ≤ 0.75), while for the non-linear parameter (Sample Entropy), it was moderate (0.19 ≤ R2 ≤ 0.29). Strongest relationships were obtained with 2-minutes windows. The findings suggest that window analysis can accurately assess contraction intensity and, more moderately its spectral content; but basal segments in window analysis significantly influence the signal complexity parameter.
Download
Paper Nr: 169
Title:

PPG Signal Quality Classification Using STFT and CNN with the BUT PPG Database
Authors:

Leandro Duque Mussio and Maria Claudia F. Castro
Abstract: Photoplethysmography (PPG) signal analysis has the potential for various medical applications, such as heart rate monitoring, blood pressure estimation, and emerging techniques like diagnosing diabetes and glucose level estimation. However, noise and artifacts, especially motion artifacts, can degrade the quality of PPG signals, making it difficult to extract meaningful features. This research addresses this challenge by investigating the quality of photoplethysmography (PPG) signals using the Short-Time Fourier Transform (STFT) and a deep learning model. The objective is to classify PPG signals as good or bad to eliminate bad signals and increase the accuracy of subsequently derived features. The signals were pre-processed using the publicly available BUT PPG database, consisting of a limited number of smartphone PPG recordings with a low sampling rate (30 Hz), generating spectrographic images used in training a Convolutional Neural Network (CNN) to classify the quality of the signals. Nested cross-validation with five external folds and two internal stratified folds was applied to optimize hyperparameters and assess the model’s performance. The results show the effectiveness of the proposed approach, improving the extraction of features from PPG signals by collecting 94.29% (± 7.82%) of good signals and filtering 80% (± 12.78%) of bad signals.
Download
Paper Nr: 174
Title:

Unified CNN-Transformer Model for Mental Workload Classification Using EEG
Authors:

Fiza Parveen and Arnav Bhavsar
Abstract: The cognitive effort required for tasks requiring memory, attention, and decision-making is referred to as mental workload. Preventing cognitive overload and increasing task efficiency rely on a reliable assessment of mental workload. In this study, we present a CNN-Transformers hybrid model that uses EEG data for multi-level Mental Workload classification. Our model uses 1D-CNN to extract spatial features from windowed EEG signals followed by Transformers to capture temporal correlation.This combination improves our comprehension of mental workload situations by capturing local spatial and both long-range temporal aspects. We use a majority voting technique on the window based predictions to increase prediction reliability, making sure the final accuracy represents a thorough assessment of mental workload at signal level. A rigorous 5-fold cross-validation technique is used to evaluate the model on publicaly available STEW dataset.
Download
Paper Nr: 191
Title:

Routine Pattern Learning and Anomaly Detection Applied to Lone Workers Through Topic Modeling
Authors:

Ana Cravidão Pereira, Marília Barandas and Hugo Gamboa
Abstract: Learning routines and identifying anomalous behaviors play a critical role in worker safety. Identifying deviations from normal patterns helps prevent accidents, ensuring enhanced safety in complex environments. Topic modeling is frequently used to discover hidden semantics patterns and is well-suited to the complexity of routines in human behavior. However, its utility in complex time-series analysis and as a baseline for anomaly detection has not been widely explored. This work proposes a novel solution to accurately model complex routines using topic modeling, enabling the identification of anomalies through a statistical approach. A dataset of human movement recordings was collected over up to seven consecutive months, capturing the routines of three lone workers, with each accumulating between 414 and 955 hours of recording time. This dataset served as the basis for a comprehensive analysis of the results, showing strong alignment between visually observed patterns in routines and the outcomes of the proposed method. Additionally, detecting anomalies across models with varying training days confirms that online learning enhances the accuracy and adaptability of routine modeling. Topic modeling allows for in-depth learning of routines, capturing latent patterns undetectable to humans. This capability prevents abnormal events, leading to a proactive approach to predictive risk assessment.
Download
Paper Nr: 207
Title:

Comparative Study of Data Processing Techniques for Pancreatic Islets in Organ-on-Chip Applications
Authors:

Roland Giraud, Dorian Chapeau, Jochen Lang, Matthieu Raoux, Sylvie Renaud and Antoine Pirog
Abstract: Organ-on-chip technology presents a promising platform to study complex physiological processes in a controlled environment. However organ-on-chip devices bring considerable constraints to online monitoring instrumentation. This study investigates methods for leveraging data from organ-on-chip systems designed for diabetic studies by processing recorded extracellular signals from pancreatic islets. The signal processing techniques used are designed to address the inherent constraints of microfluidics, particularly to provide online (real-time) readings and operate effectively in low Signal-to-Noise Ratio (SNR) conditions. This study assesses the performance of different algorithms using several detection approaches. Synthetic and experimental data were utilized to evaluate algorithm robustness to best account for biological variability. Among the algorithms tested, those based on frequency and time-frequency methods performed best when compared to conventional filtering and thresholding approaches, especially regarding robustness to noise and biological variability.
Download
Paper Nr: 210
Title:

Behavior-Based Deepfake Detection: Leveraging Cognitive Response to Visual Face Perception
Authors:

Hendrik Graupner, Mohammad Yeghaneh Abkenar, Lisa Schwetlick, Ralf Engbert and Christoph Meinel
Abstract: Face presentation attacks are a propagating issue in an increasingly digitally interconnected world. One of the most recent developments is deepfake impersonation attacks in live video streams. Behavioral biometric analysis is a crucial part of a comprehensive solution to this pressing issue. This paper proposes the application of biological responses to visual self-recognition as a dynamic biometric trait. Self-recognition is a cognitive process that can be leveraged as in-brain identity validation. A sophisticated pre-trained model classifies eye-tracking data to determine the face in the user’s current visual focus. One eminent use case is the protection of online video conferences. This paper provides the architecture of a prototypical implementation based on an open-source video conferencing platform. Our work of interdisciplinary research aims to contribute to a holistic solution to protect our modern communication systems and restore trust in digitization.
Download
Paper Nr: 218
Title:

Enhanced Assessment of Gait Dynamics in Multiple Sclerosis: A Signal Processing Approach for Extracting Range of Motion Using Wearable IMUs
Authors:

Dunya Moradi, Kardelen Akar, Hussein Youssef, Ayse Altintas, Atay Vural and Beren Semiz
Abstract: This study investigates the gait dynamics and motor impairment severity in individuals with multiple sclerosis (MS) by analyzing lower limb range of motion (ROM) using data from inertial measurement units (IMUs) collected during the Timed 25-Foot Walk (T25FW) clinical task. Forty-eight participants were categorized into two MS groups based on motor impairment severity (16 MS patients with low motor impairment, 16 MS patients with moderate to severe motor impairment) and 16 healthy control group. IMU raw data of accelerometer and gyroscope from the feet sensors with respect to the lumbar region, were processed using a Butterworth filter and an Extended Kalman Filter to obtain accurate orientation, followed by quaternion to Euler angle conversion for calculating ROM. When the ROM-extracted statistical and time domain features were compared, there were significant differences in ROM characteristics among groups, particularly highlighting the increased variability and reduced control in participants with severe motor impairments. ROM-extracted features such as kurtosis, skewness, and entropy underscored the asymmetrical and irregular motion patterns in advanced MS cases. These findings support the potential of IMU-derived ROM metrics as biomarkers for tracking MS disease progression and tailoring rehabilitation.
Download
Paper Nr: 228
Title:

Toward Designing a Reduced Phone Set Using Text Decoding Accuracy Estimates in Speech BCI
Authors:

Shuji Komeiji, Koichi Shinoda and Toshihisa Tanaka
Abstract: Reducing the phone set in speech recognition or speech brain-computer interface (BCI) tasks improves phone discrimination accuracy. This reduction may also degrade text decoding accuracy due to increased homonyms. To address this, we propose a novel estimator called the Generalized Pronunciation/Word Confusion Rate (GPWCR), which estimates text decoding accuracy by considering both phone discrimination performance and the number of homonyms. By minimizing the GPWCR, we designed the optimal reduced phone set. Experimental results from Japanese large vocabulary speech recognition demonstrate that the optimal phone set, reduced from 39 to 38 phones, lowered the word error rate from 14.1% to 13.8%.
Download
Paper Nr: 229
Title:

Effects of Class Imbalance in Unsupervised Human Activity Recognition for Office Work Task Characterization
Authors:

Sara Santos, Phillip Probst, Luís Silva and Hugo Gamboa
Abstract: Office workers spend most of their time sitting, often with rigid postures, for prolonged periods of time. This has been recognized by the European Union as a risk factor for work-related musculoskeletal disorders. To study work activities and their distribution over time, Human Activity Recognition (HAR) techniques need to be implemented. Since supervised learning techniques require labeled data and large datasets for training, unsupervised learning is a viable alternative for HAR. However, these models may be affected by the highly imbalanced distribution of activities typically observed in office workers. Considering this, this work studied the impact of data imbalance on clustering performance when the dataset is comprised of 33 %, 50 %, 70 %, and 90 % of sitting activity. Office activities were collected from 19 subjects and three traditional clustering models were employed. KMeans and Gaussian Mixture Model were more affected than Agglomerative Clustering, which seems to be more robust to data imbalance. With 90 % of sitting time, all three models performed poorly, which emphasizes the need for clustering models that can handle highly imbalanced data.
Download
Paper Nr: 230
Title:

Review on the Effects of Hypergravity on Workload and Fine Motor Skills in Humans
Authors:

Judith Bütefür, Julia Habenicht and Elsa Andrea Kirchner
Abstract: As the human body has adapted all functions and movements to Earth gravity it has to adapt its functions and movements when gravitational forces increase. Astronauts are for example exposed to increased gravity, i.e., hypergravity, during rocket launches. To prevent security incidents on the missions, it is important to achieve the best possible cognitive and motor performance from the outset, which requires a better understanding of cognition and behavior under hypergravity. The aim of this paper is to provide an overview of studies investigating the electroencephalogram (EEG), the electrocardiogram (ECG), muscle activity (EMG) and aiming accuracies in hypergravity as these biosignals are known to capture human cognitive performance and motor performance in order to make a statement about the effect of hypergravity on the human body. The literature review shows that all investigated parameters change under hypergravity. This fact highlights the need for further analysis of how these changes affect the human body in relation to motor performance. It also shows the need for novel and flexible training methods that allow astronauts to acclimatize to the new gravity conditions without limiting the duration of training, such as when training takes place during a parabolic flight or using a centrifuge. We propose the use of an active upper body exoskeleton as a new and flexile training method.
Download
Paper Nr: 246
Title:

An EEG-Based Study Investigating Cognitive and Behavioral Reactions to Indian Knowledge System Narratives in Virtual Reality
Authors:

Sakshi Chauhan, Gitanshu Choudhary, Arnav Bhavsar and Varun Dutt
Abstract: Virtual reality (VR) storytelling has been demonstrated to impact emotional and cognitive capacities. Still, less is known about the precise impacts of VR on moral learning and engagement, especially when it comes to Indian Knowledge System (IKS) stories. In order to close this gap, this study examines the moral learning and engagement potential of VR-based storytelling in comparison to traditional reading. Three groups of 75 individuals each were assigned to VR, reading, and control. The VR and reading groups outperformed the control group in terms of moral learning and retention, according to behavioral data, which did not reveal any significant differences between them. However, according to the EEG data, the VR group was more engaged than the reading group, as evidenced by a lower alpha band power. Participants using VR showed higher engagement, as evidenced by 88% of responses indicating agreement or strong agreement on a five-point Likert scale. These results imply that while reading and VR are equally helpful for moral learning, VR is more engaging due to its immersive features.
Download
Paper Nr: 269
Title:

Reproducible Gating for High-Resolution Flow Cytometric Characterization of Extracellular Vesicles in Next-Generation Biomarker Studies
Authors:

Ishwor Thapa, Yohan Kim, Fabrice Lucien and Hesham Ali
Abstract: With the continuous advancements of biomedical technologies, we have access to instruments capable of producing new types of biological data or generating traditional data with higher degrees of quality. With the support of such data, researchers and practitioners continue to explore the possibilities of developing new approaches to obtain valuable data-driven signatures or biosignals to be used for diagnosis, classification, or assessment of treatments. However, with the emergence of new types of data, it is often the case that they are available in raw formats that are not suitable for extracting the needed biomarkers. Hence, much work is needed to process the raw data sets obtained from new medical instruments and transform the signals into products capable of capturing the desired knowledge. Next-generation biomarkers such as “liquid biopsies” are emerging tools to improve cancer diagnostics, disease stratification, and treatment monitoring. As potential cancer biomarkers, circulating Extracellular Vesicles (EV) levels may early-predict disease recurrence and resistance to treatment. High-resolution flow cytometry (hrFC) is a sensitive and high-throughput method for quantifying circulating levels of EVs with minimal sample processing. One of the benefits of using hrFC is that there is no need to isolate or purify the molecules of interest from the biological samples prior to running the flow. However, signals in hrFC data currently depend on manual and subjective approaches to gating the positive events. Such approaches are often time-consuming, error-prone, and lack the levels of robustness and reproducibility needed to trust the obtained information. This study proposes an automated quantitative technique to process flow cytometry data for EVs with a high degree of accuracy consistency. A publicly available Shiny web application is presented that performs quality check of flow cytometry files and automated gating of biosignals, viz. subpopulations of EVs that are of interest to next generation biomarker studies.
Download
Paper Nr: 297
Title:

Facial Profile Biometrics: Domain Adaptation and Deep Learning Approaches
Authors:

Malak Alamri and Sasan Mahmoodi
Abstract: Previous studies indicate that human facial profiles are considered as a biometric modality and there is a bilateral symmetry in facial profile biometrics. This study examines the bilateral symmetry of the human face profiles and presents the analysis of facial profile images for recognition. A method from few-shot learning framework is proposed here to extract facial profile features. Based on domain adaptation and reverse validation, we introduce a technique known as reverse learning (RL) in this paper for the same side profiles to achieve a recognition rate of 85%. In addition, to investigate bilateral symmetry, our reverse learning model is trained and validated on the left side face profiles to measure the cross recognition of 71% for right side face profiles. Also in this paper, we assume that the right face profiles are unlabelled, and we therefore apply our reverse learning method to include the right face profiles in the validation stage to improve the performance of our algorithm for opposite side recognition. Our numerical experiments indicate an accuracy of 84.5% for cross recognition which, to the best of our knowledge, demonstrates higher performance than the state-of-the-art methods for datasets with similar number of subjects. Our algorithm based on few-shot learning can achieve high accuracies for a dataset characterized with as low as four samples per group.
Download
Paper Nr: 332
Title:

Predicting Respiratory Depression in Neonates Using Deep Learning Neural Networks
Authors:

Aleksandar Jeremic and Dejan Nikolic
Abstract: Respiratory problems are one of the most common reasons for neonatal intensive care unit (NICU) admission of newborns. It has been estimated that as much as 29% of late preterm infants develop high respiratory morbidity. To this purpose invasive ventilation is often necessary for their treatment in NICU. These patients usually have underdeveloped respiratory system with deficiencies such as small airway caliber, few collateral airways, compliant chest wall, poor airway stability, and low functional residual capacity. Consequently ventilation control has been subject of considerable research interest. In this paper we propose an algorithm for detection of respiratory depression by predicting the onset of pO2 depressions using physiological measurements. We train deep neural network using previously obtained data set from NICU, McMaster University Hospital with intra-arterial pressure measurements and evaluate its performance. Preliminary results indicate that adequate performance can be achieved if sufficient number of measurements is available.
Download
Paper Nr: 44
Title:

Changes in Attention Levels While Driving a Car Estimated Using Modelling Techniques with Features of Oculo-Motors
Authors:

Minoru Nakayama, Qian Chayn Sun and Jianhong Cecilia Xia
Abstract: Changes in attention levels while driving a car were estimated using a modelling technique involving pupillary changes and the frequency of saccades of 11 drivers. The driving route used in the experiment consisted of 19 sections of road divided into 5 groups: university campus, left turn, straight, right turn, and roundabout. The sections of road with posted speed limits were divided into 6 conditional states, and model parameters were estimated by assuming transitions across the states. The estimated model parameters were used to examine changes in the level of attention resources used during each section of driving. The results of a comparison of attention resources by section showed a significant decrease, in the following order: straight and roundabout, within campus, left turn and right turn. In addition, the relationship between NASA-TLX was evaluated after driving and attention resources were examined, and a significant correlation with the factor for “difficulty” was confirmed. The relationship between the confidence interval of the change in attention resources and the factor for “mental demand” was also confirmed.
Download
Paper Nr: 108
Title:

Generative Adversarial Network for Image Reconstruction from Human Brain Activity
Authors:

Tim Tanner and Vassilis Cutsuridis
Abstract: Decoding of brain activity with machine learning has enabled the reconstruction of thoughts, memories and dreams. In this study, we designed a methodology for reconstructing visual stimuli (digits) from human brain activity recorded during passive visual viewing. Using the MindBigData EEG dataset, we preprocessed the signals and cleaned them from noise, muscular artifacts and eye blinks. Using the Common Average Reference (CAR) method and past studies’ results we reduced the available electrodes from 14 to 4 keeping only those containing discriminative features associated with the visual stimulus. A convolutional neural network (CNN) was then trained to encode the signals and classify the images. A 92% classification performance was achieved post-CAR. Three variations of an auxiliary conditional generative adversarial network (AC-GAN) were evaluated for decoding the latent feature vector with its class embedding and generating black-and-white images of digits. Our objective was to create an image similar to the presented stimulus through the previously trained GANs. An average 65% reconstruction score was achieved by the AC-GAN without a modulation layer, a 60% by the AC-GAN with modulation layer and multiplication, and a 63% by the AC-GAN with modulation and concatenation. Rapid advances in generative modeling promise further improvements in reconstruction performance.
Download
Paper Nr: 123
Title:

Analyzing Male Depression Using Empirical Mode Decomposition
Authors:

Xavier Sánchez Corrales, Jordi Solé-Casals, Enrique Arroyo García and Diego Palao Vidal
Abstract: This study investigates the differences in male voice between healthy individuals and individuals with depression, using Empirical Mode Decomposition (EMD) analysis. Voice recordings from 25 men with depression and 76 without were analyzed. The methodology consisted of extracting 16 Intrinsic Mode Functions (IMFs) from 20-second voice segments, followed by statistical analyses including bootstrapping of means and standard deviations with False Discovery Rate (FDR) correction, comparison of probability density functions, and the application of a Gaussian kernel. The results showed significant differences between the means and standard deviations, The application of the Gaussian kernel revealed more pronounced differences in IMFs 2 to 6, providing more specific discrimination than traditional statistical methods. The study contributes to the development of non-invasive and objective diagnostic tools for depression.
Download
Paper Nr: 143
Title:

Comparison Between the Effects of Continuous and non-Continuous Visual Feedback on Motor Learning While Playing a Muscle-Controlled Serious Game
Authors:

Julia Habenicht and Elsa Andrea Kirchner
Abstract: The guidance hypothesis suggests that continuous feedback during learning may lead to feedback dependency, with errors decreasing when feedback is provided and increasing when it is removed. This study investigates the effect of continuous (CVF) versus non-continuous visual feedback (NCVF) on motor learning using a muscle- controlled serious game. Subjects played the game for three consecutive days, with each day consisting of seven training sets and one learning control set without feedback. One group received CVF during training, while the other received NCVF. To assess transferability, the results of the learning control sets were compared between groups. Time to success during training decreased for CVF, and average time to reach the longest correct time period in the learning control set was higher for CVF compared to NCVF. The number of missed goals decreased for CVF, aligning with the expected positive impact of continuous feedback during training. However, the results for the learning control sets were inconclusive. While CVF showed a potential dependency on feedback, the decrease in missed goals indicates improved motor learning. More test days and subjects are required to confirm the findings and draw definitive conclusions regarding the guidance hypothesis.
Download
Paper Nr: 195
Title:

Development of a Procedure for Detecting Dementia Symptoms Using Features in Differential Waveforms of the Pupil Light Reflex
Authors:

Minoru Nakayama, Wioletta Nowak and Anna Żarowska
Abstract: A procedure for detecting dementia levels is developed using features of waveform shapes of the pupillary light reflex (PLR) in response to chromatic light pulses on either eye. Features of waveform shapes were extracted using a functional data analysis technique which measured the reactions of both eyes. In considering the physiological mechanism, differential waveform shapes were also analysed. The feature was extracted as a coefficient of B-spline basis functions of the waveforms. The feature sets of the differential waveform shapes for blue and red light pulses contributed to detection performance. Also, feature weights are used to represent PLR reaction mechanisms and differences in response to chromatic stimuli.
Download
Paper Nr: 202
Title:

Investigating Behavioral and Neurophysiological Responses Across Landslide Scenarios in Virtual Reality
Authors:

Arjun Mehra, Arti Devi, Ananya Sharma, Sahil Rana, Shivam Kumar, K. V. Uday and Varun Dutt
Abstract: The potential of virtual reality for disaster preparedness is enormous, but little is known about how different landslide risks and environmental factors (day versus night) affect human reactions. Neurophysiological (alpha/theta, alpha/gamma, and beta/gamma ratios from EEG) and behavioral (Euclidean distance, collisions, and velocity around collisions) measures are combined in this study to investigate stress and cognitive engagement in landslide simulations. In order to expose 80 participants to varying landslide probabilities, they were randomly assigned to four groups with varying landslide risk and lighting conditions. Behavioral deviations and cognitive workload were significantly influenced by perceived risk rather than lighting conditions, according to the results. Electroencephalography (EEG) and behavioral outcomes were correlated, which emphasized how crucial integral analysis is to comprehending disaster responses. These results demonstrate how well virtual reality can develop cognitive resilience and offer guidance for creating training plans that maximize performance in high-risk scenarios. This study develops dynamic, immersive VR-based disaster preparedness apps.
Download
Paper Nr: 235
Title:

Towards Human Posture Detection Based on Differential Measurements Using Wearable Barometric Pressure Sensors
Authors:

Nico Graumüller, Constantin Gis, Franziska Geiger, Iman Soodmand, Maeruan Kebbach, Rainer Bader, Christian Haubelt and Florian Grützmacher
Abstract: The detection of human postures is a well-studied research area that is closely related to human activity recognition. Recent advantages of MEMS-based barometric pressure sensors have made them an interesting additional sensing modality apart from IMU-based approaches. State-of-the-art barometric pressure sensors allow for measuring changes in barometric pressure corresponding to height differences in the range of centimeters. However, they are susceptible to environmental pressure changes, which can significantly influence the application. Therefore, we propose a posture detection approach based on differential height measurements from multiple body-worn barometric pressure sensors. We conducted an initial laboratory study with 13 subjects (eight males and four females), evaluating standing, sitting, and lying down postures using four body-worn barometric pressure sensors positioned at the head, hip, wrist, and ankle. Our results demonstrate that only two sensors are needed to separate the studied postures in the feature space. Furthermore, the differential height measurement approach can compensate for environmental pressure influences to an insignificant level w.r.t. posture separability in our setup. The efficacy of our proposed approach is further substantiated by the observed separability of sitting on a bed and a chair for each subject individually.
Download
Paper Nr: 283
Title:

A Novel Approach to Modelling Multi-Channel and Multi-Phase Signals on an Angular Coordinate Axis (Semani)
Authors:

Selma Ozaydin
Abstract: This study introduces a new digital signal processing method that holistically plots digital signal vectors collected from multiple channels on an angular graphic axis and aims to obtain a holistic signal matrix in the specified angular range using a vector interpolation technique. The proposed method, Semani, enables the visualization, improvement, and analysis of signal parameters based on phase angles and independent variables using an angular coordinate axis. In the Semani method, multi-channel signals are visualized on a single graph by plotting them on a coordinate axis encompassing all angular directions. This approach divides the angular coordinate axis into sections based on its resolution level of angles (segments) and the rate of change of independent variables (layers) within the analysis window. A graphic pattern (polar, cartesian, cylinder, sphere, etc.) determined on the angular axis is divided into slices according to segments and layers, and the signal is plotted in these sections. The proposed signal plotting and analysis model enables holistic modeling of multi-channel signals collected from different angular directions on a coordinate axis. Additionally, the vector interpolation method used in this model calculates signal vectors for unknown angular directions, enriching the signal. This innovative method allows signals collected from multiple channels, such as EEG, ECG, radar, sonar, and seismic signals, to be effectively visualized on a single graph against their corresponding independent variables (e.g., time, frequency, distance).
Download
Paper Nr: 286
Title:

Heart Rate Turbulence: Wavelet Analysis of Frequency Modulated Signals
Authors:

S. V. Bozhokin, I. B. Suslova, A. A. Riabokon and T. D. Shokhin
Abstract: This paper presents new approaches to the analysis of non-stationary heart rate variability (HRV) taking into account strong time variation in the duration of RR intervals, which is associated with extrasystoles. A mathematical model of a frequency-modulated signal comprising of identical Gaussian peaks unevenly spaced along the time axis is applied to the phenomenon of heart rate turbulence (HRT). The maxima of the Gaussian peaks correspond to the moments of real heart contractions. A time-continuous function of local (instantaneous) heart rate frequency is calculated by analyzing the maxima of the continuous wavelet transform applied to such a model signal. The change in local frequency over time is proposed as a new characteristic of extrasystoles and compensatory pauses in the heart tachogram. The proposed method, applied in this work to study tachogram records with extrasystoles, can be used in the analysis of any other heart rhythm disturbances.
Download