[STAGE MASTER] - Explainable Feature Selection for Stress and Emotion Monitoring Systems Using the WESAD Dataset

Scientific fields: Artificial Intelligence, Machine Learning, Feature Selection, Digital Health. Keywords: Feature Selection, Multimodal Data, Real-Time, Monitoring, Fall-detection, Health. Internship Topic Stress and emotion monitoring systems play a crucial role in mental health management, especially given the global rise in psychological disorders. Wearable devices equipped with sensors such as ECG (electrocardiogram), PPG (photoplethysmogram), EDA (electrodermal activity), and motion sensors provide a rich source of data for stress and emotion analysis. However, the high dimensionality and noise present in these multimodal datasets pose challenges for efficient processing and interpretability. Feature selection is essential for building robust and interpretable models. Recent advancements in explainable AI (XAI) emphasize the need to ensure that selected features provide understandable information about emotional and stress states, which strengthens trust and adoption by healthcare professionals. The WESAD (Wearable Stress and Affect Detection) dataset offers an ideal platform to develop and test these approaches. It includes multimodal physiological and motion data collected during controlled experiments, labeled for different stress and emotional states. Internship Objective Develop an explainable feature selection framework tailored to multimodal physiological data, enabling precise and interpretable monitoring of stress and emotions using the WESAD dataset. Research Objectives Feature Selection Framework: Design a feature selection method to identify the most relevant features in physiological signals (ECG, EDA) and motion data (accelerometer). Ensure robustness against noise and variability in physiological data. Explainability and Interpretability: Integrate explainable AI techniques (e.g., SHAP, LIME) to explain the importance of selected features in detecting stress and emotional states. Real-Time Application: Optimize the framework for low-latency applications suitable for real-time stress monitoring. Methodology: 1. Data Preprocessing: Extract and preprocess signals from the WESAD dataset (ECG, EDA, PPG, accelerometer). Synchronize modalities and address issues related to noise and missing data. 2. Feature Extraction and Selection: Extract domain-specific features (e.g., heart rate variability, signal entropy, motion patterns). Develop a hybrid feature selection framework: Filters: Use statistical measures like correlation and mutual information. Wrappers: Implement recursive or sequential feature selection. Embedded Methods: Leverage tree-based models (e.g., Random Forest) to analyze feature importance. 3. Explainable AI Integration: Use SHAP or LIME to evaluate and visualize the contribution of selected features to model predictions. Ensure interpretability aligns with clinical relevance (e.g., heart rate variability for stress detection). 4. Model Development and Validation: Train machine learning models (e.g., SVM, LSTM, or CNN) using selected features. Evaluate performance using metrics such as: Classification accuracy: precision, F1-score. Real-time efficiency: latency, resource consumption. Explainability: variable importance analysis. Compare results with state-of-the-art methods in stress monitoring. Expected Outcomes: 1. - A robust and explainable framework for feature selection in stress and emotion monitoring. 2. - Improved accuracy and efficiency in stress detection models. 3. - Interpretable insights into physiological and motion indicators of stress, contributing to mental health research. 4. - Publication in a journal or conference focused on AI or health. Introduction to the laboratory CESI LINEACT- Research Unit CESI LINEACT (Digital Innovation Laboratory for Companies and Learnings at the service of the territories competitiveness) is the CESI group laboratory whose activities are implemented on CESI campuses. Link to the laboratory website: https://lineact.cesi.fr/en/https://lineact.cesi.fr/en/research-unit/presentation-lineact/ CESI LINEACT (EA 7527), Digital Innovation Laboratory for Business and Learning at the service of the Competitiveness of Territories, anticipates and accompanies the technological mutations of the sectors and services related to industry and construction. CESI's historical proximity to companies is a determining factor for our research activities and has led us to focus our efforts on applied research close to companies and in partnership with them. A human-centered approach coupled with the use of technologies, as well as the territorial network and the links with training, have allowed us to build transversal research; it puts the human being, his needs and his uses, at the center of its problems and approaches the technological angle through these contributions. Its research is organized according to two interdisciplinary scientific themes and two application areas. Theme 1 "Learning and Innovation" is mainly concerned with Cognitive Sciences, Social Sciences and Management Sciences, Training Sciences and Techniques and Innovation Sciences. The main scientific objectives of this theme are to understand the effects of the environment, and more particularly of situations instrumented by technical objects (platforms, prototyping workshops, immersive systems, etc.) on the learning, creativity and innovation processes. Theme 2 "Engineering and Digital Tools" is mainly concerned with Digital Sciences and Engineering. The main scientific objectives of this theme concern the modeling, simulation, optimization and data analysis of industrial or urban systems. The research work also focuses on the associated decision support tools and on the study of digital twins coupled with virtual or augmented environments. Research intersects across the application domains of the Factory of the Future and the City of the Future. References: 1. chmidt, P., Reiss, A., Duerichen, R., Marberger, C., & Van Laerhoven, K. (2018). WESAD: A Multimodal Dataset for Wearable Stress and Affect Detection. Proceedings of the 20th ACM International Conference on Multimodal Interaction. 2. Lundberg, S. M., & Lee, S.-I. (2017). A Unified Approach to Interpreting Model Predictions. Advances in Neural Information Processing Systems. 3. Healey, J., & Picard, R. W. (2005). Detecting Stress During Real-World Driving Tasks Using Physiological Sensors. IEEE Transactions on Intelligent Transportation Systems.

Location: Villeurbanne, FR

Posted Date: 2/5/2025