Sophisticated bio-sensing architecture for anxiety identification using enhanced machine learning techniques

Shivendra Dubey, Sakshi Dubey, Sachin Malviya, Nandkishor Joshi, Pranshu Pranjal,
Sophisticated bio-sensing architecture for anxiety identification using enhanced machine learning techniques,
Medicine in Novel Technology and Devices,
Volume 28,
2025,
100405,
ISSN 2590-0935,
https://doi.org/10.1016/j.medntd.2025.100405.
(https://www.sciencedirect.com/science/article/pii/S2590093525000566)
Abstract: In recent years, medical wearable technologies are growing more and more popular as mental awareness and psychological peace have grown in importance. With the help of this modern technology, continuous surveillance is made possible, giving healthcare workers vital physiologic information that can improve the treatment of patients. The noise's susceptibility and rigidity interference vulnerability impact on present anxiety-detection techniques, like the electrocardiogram (ECG), blood volume pulse (BVP), and movement analysis of the body are the main techniques. We present an ANXIETY-CARE, a flexible anxiety detection system that uses a hybrid methodology to get around these constraints. This smart mechanism makes use of machine learning techniques, state-of-the-art context detection techniques, and a sweat sensor device. The ANXIETY-CARE uses the XG Boost classification algorithm to process sensor model as well as data variations in the natural world. We hope to transform anxiety detection by fusing those cutting-edge methods, providing an additional flexible and reliable option for general wellbeing and better anxiety management. By utilizing Electrodermal Response (EDR) sweat sensors, we present a cutting-edge anxiety detection tool in the suggested approach that outperforms conventional Electrocardiogram (ECG) techniques without requiring implants. This study clarifies noisy contextual interpretation for diverse wearing equipment, providing essential guidelines for improving anxiety identification over multiple applications and contexts. We are incorporating machine learning techniques, specifically XG-Boost techniques for improvement of diagnosis reliability and accuracy. In this study, we also forecast anxiety employing handheld data from sensors, and also investigated the use of CNN-LSTM, CNN, and LSTM models. These models deal directly with data that occurs over time, in contrast to earlier approaches that depended on stable properties. We identified LSTM approach demonstrated potential for subsequent studies with big datasets, in this study we have used StudentLife dataset and achieving 84.3 ​% accuracy.
Keywords: ANXIETY; XG boost; Machine learning; ECG; EDR