Language：English   Publishing type：Research paper (international conference proceedings)

Aalto University, Espoo, Finland   https://www.ubicomp.org/ubicomp-iswc-2025/   2025.10.14  -  2025.10.16

In this paper, team VedaNam Coders proposes a Reservoir Computing approach to address robustness and generalization in the 2nd WEAR Dataset Challenge of UbiComp HASCA 2025. The main task involves recognising 18 activities from inertial data collected at four locations in the body. The challenge emphasises generalisation and robustness to randomly sampled, sensor-specific 1-second windows from unseen participants, with augmentations simulating diverse wearing conditions. Prior baseline approaches, such as DeepConvLSTM and TinyHAR, struggle with low-motion or overlapping classes with limitations on handling short windows, and noisy inputs. In response, we propose a two-stage hybrid pipeline combining CNNs for spatial feature learning and Reservoir Computing (RC) for short-term temporal dynamics. To evaluate robustness and generalization, we compared our RC-based approach with three gradient boosting models and evaluated four RC variations. Experimental results show that the merged-limb two-stage RC strategy achieves the highest test macro F1 score (0.52888), outperforming the three challenge baselines and tree-based models. This study demonstrates the potential of reservoir computing across different granularities and architectures for building robust HAR models.

DOI： https://doi.org/10.1145/3714394.3756190

DOI： https://doi.org/10.1145/3714394.3756190

Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

Kitakyushu, Japan   https://cennser.org/ICIEV/index.html   2025.05.26  -  2025.05.29

n this paper, we aim to develop a stress detection model for caregivers using data collected from a real-world care facility. The importance of stress management has grown, and many physiological datasets and stress detection studies exist for nurses. Caregivers experience various types of stress, including interpersonal issues and assisting with toileting, and are physically active throughout the day, moving around the floor and performing tasks like transfer assistance. These diverse stressors and physical activity can significantly affect stress detection using wearable sensors. However, sufficient data collection in caregiving settings is not
being conducted compared to other fields (e.g., nursing, doctor). In this study, we conduct a data collection experiment and evaluate the results of stress detection machine learning using that data. In the data collection, we create 8-day dataset from four caregivers that includes care record data, wearable sensor data, and self-reported stress labels. Next, we extracted statistical features from the time-series data and performed stress detection using a Random Forest model. As a result, we achieved a maximum classification accuracy of 72%. While data augmentation improved the detection of minority classes such as High Stress, it also lowered the overall classification performance, revealing a trade-off that remains a challenge. Nevertheless, this dataset makes it possible to analyze caregiver stress in relation to specific care activities, representing an important step toward stress-aware support in the caregiving domain.

DOI： https://doi.org/10.60401/ijabc.120

DOI： https://doi.org/10.60401/ijabc.120

Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

Abu Dhabi, United Arab Emirates   https://abc-research.github.io/2025   2025.04.21  -  2025.04.25

The Silent Speech Decoding Challenge, hosted as part of the ABC 2025 conference, was designed to advance the decoding of silent (inner) speech from electroencephalography (EEG) signals, specifically targeting Arabic language processing. Participants leveraged advanced machine learning (ML) and deep learning (DL) techniques to classify EEG data from a uniquely collected dataset. This dataset, gathered by Qatar University's Machine Learning Group, consisted of EEG recordings from ten subjects across multiple sessions, capturing both silent and overt speech for six distinct inner speech commands. Participants were tasked with developing robust ML/DL models to accurately decode these neural signals, addressing challenges such as generalization across subjects and robustness to noise and motion artifacts.

DOI： https://doi.org/10.60401/ijabc.104

DOI： https://doi.org/10.60401/ijabc.104

Other Link： https://autocare.ai/abc2025

Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

Abu Dhabi, United Arab Emirates   https://abc-research.github.io/2025   2025.04.21  -  2025.04.25

Early detection of depressive episodes is crucial in managing mental health disorders such as Major Depressive Disorder (MDD) and Bipolar Disorder. However, existing methods often necessitate active participation or are confined to clinical settings. While facial expressions have shown promise in laboratory settings for identifying depression, their potential in real-world applications remains largely unexplored. In this challenge we introduce a data collected state of the art facial behavior sensing system, that tracks different facial behavior primitives such as Action Units, Landmarks, Head Pose, Eye Open state and others for task of detecting depressive episode in the wild. The challenge duration was three months, from Dec 12, 2024 to Feb 28, 2025 and 8 teams participated. Among the final teams, Team Persistence (Team ID 160) achieved 0.77 AUROC for universal model and 0.88 AUROC for hybrid model and became the winner.

DOI： https://doi.org/10.60401/ijabc.102

DOI： https://doi.org/10.60401/ijabc.102

Other Link： https://autocare.ai/abc2025

Authorship：Corresponding author   Language：English   Publishing type：Research paper (international conference proceedings)

Abu Dhabi, United Arab Emirates   https://abc-research.github.io/2025   2025.04.21  -  2025.04.25

Recognizing normal and unusual activities (e.g., tremor) from sensor data, particularly those related to the wear-off phenomenon of Parkinson’s disease (PD), is crucial to improving the quality of life of elderly individuals. This paper presents five selected methods from the Activity and Behavior Computing Tremor Challenge 2025, which focused on developing robust models to recognize Parkinson’s disease-related activities using wearable triaxial accelerometer data. These approaches range from traditional machine learning pipelines based on handcrafted multi-domain features to advanced deep learning architectures incorporating temporal modeling and attention mechanisms.
The review also highlights innovative strategies such as adaptive window segmentation and data augmentation using conditional variational autoencoders. Each method addressed key challenges, including data label misalignment, class imbalance, noisy sensor readings, and the limited size of the dataset. The explored models ranged from traditional XGBoost classifiers to deep learning architectures such as CNN-LSTM hybrids, attention-enhanced DeepConvLSTM and Transformer-based classifiers, each demonstrating varying levels of
success. This summary provides insights into the trade-offs between model interpretability and predictive performance, emphasizes the effectiveness of combining feature engineering with deep learning, and outlines directions for future research in scalable PD monitoring systems. The findings highlight the need for more advanced approaches to reliably decipher these challenging activities.

DOI： https://doi.org/10.60401/ijabc.103

DOI： https://doi.org/10.60401/ijabc.103

Other Link： https://autocare.ai/abc2025

Event date： 2025.04.21 - 2025.04.25   Language：English   Country：United Arab Emirates  

In this study, we use Large Language Models (LLMs) for room recognition in nursing facilities, leveraging contextual data to address data limitations. Traditional models struggle with Received Signal Strength Indicator (RSSI) variability and sparse data. LLMs offer a promising alternative by integrating domain knowledge to address signal inconsistencies and improve localization accuracy.

Other Link： https://abc-research.github.io/2025

Country：Japan

Awarded Best Short Paper for the work "LLM-based Indoor Localization in Nursing Care Facility: A Zero-Shot Ablation Study" presented during the 12th International Conference on Informatics, Electronics & Vision (ICIEV) in Japan last May 26-29, 2025.

Country：United Arab Emirates

Awarded with Outstanding Contribution for actively serving as Challenge Organization Chair co-hosting four coding challenges with various universities while also serving as the Publication Chair.

Country：Germany

Awarded Excellent Paper for the work "A Relabeling Approach to Signal Patterns for Beacon-based Indoor Localization in Nursing Care Facility" presented in the 5th International Conference on Activity and Behavior Computing held in Kaiserslautern, Germany last September 7th - 9th, 2023.

Country：Singapore

Awarded Best Paper for the work titled "Non-invasive Diabetes Detection using Facial Texture Features Captured in a Less Restrictive Environment" presented during the 22nd World Conference on Applied Science, Engineering and Technology held on September 2019 in Singapore.

Country：Philippines

Bachelor thesis implemented using Synopsis Industry tool grade for design. Frequency Detector with NOR gates and divide-by-64 with pseudo-NMOS divide-by-2 frequency divider is proposed, designed and simulated in TSMC 0.18 um 1P6M CMOS process technology to come up with minimum chip area and achieve fast lock-in time. This PLL design is specifically intended for Continuous-Time Sigma-Delta ADC operating at 640MHz frequency which is an important component of ICs used in electronics and communication devices whose clock rates and timing relationships are vital. This work has a lock-time of around 2.5 us which is a fast lock-in value for the lock-in time of ADC clock generator. T

Audience： College students,　Graduate students,　Teachers,　Researchesrs,　Scientific,　Governmental agency

Type：Seminar, workshop

I delivered a talk on designing Human-Centered AI solutions to indoor mobility systems and how to collaborate with stakeholders for a successful implementation of the project.

Audience： College students

Type：Seminar, workshop

Delivered a talk on how to do research abroad as international student and how to conduct multi-disciplinary research collaborations.

Audience： College students

Type：Lecture

I gave a lecture on how to conduct multi disciplinary research on the application of Artificial Intelligence to Healthcare covering 3 case studies of research works from image, video and sensor data.

Audience： College students

Type：Seminar, workshop

Delivered a talk on how to do research abroad as international student and how to conduct multi-disciplinary research collaborations.

Audience： College students,　Graduate students,　Teachers,　Researchesrs

Type：Seminar, workshop

Delivered a talk on how to apply for research scholarships abroad for graduate school and how to succeed during the study to graduate on time. What to prepare, what to expect and how to adjust living as a foreigner.