IoT-enabled Musical Therapy to Alleviate Physiological Stress in College Students using Big Data and Mixed-Density Neural Networks

Abstract

In recent years, the Internet of Things (IoT), Machine Learning (ML), and Big Data (BD) technologies have played important roles in progressing healthcare and stress management solutions. The technology allows for constant supervision of patients’ conditions, immediate data analysis, and individualized treatment courses by improving healthcare effectiveness in treating numerous health challenges. When examining physiological stress in college students, the stress level can influence students’ results and well-being. Given these challenges, this paper proposed a new IoT-based system utilizing ML and BD techniques, specifically the Mixed-Density Neural Networks (MDNN) technique, for stress improvement through musical therapy. The proposed MDNN incorporates several neural network structures to perform and analyze numerous input signals by making it individualized and consistently delivering therapeutic music. The suggested study commences by compiling various datasets involving data from microphones, physiological signals, and the environment, as these datasets are crucial for developing a holistic approach that understands and eradicates stress through music therapy. After that, the proposed work examines other methods used in feature extraction to process and analyze this data, which is vital in improving the performance of the MDNN model. The suggested MDNN employs several neural network structures to process the multi-modal inputs by allowing the real-time adjustment of therapeutic music based on the user’s stress level. Experimental results highlight the MDNN’s impressive performance metrics: accuracy, sensitivity, specificity precision, F1-score, and MCC 90.38%, 91.20%, 89.50%, 88.75%, 89.95%, and 0.82%, respectively. Moreover, the results show minimal error metrics with MAS RMSE Huber Loss and MAE, 0.15, 0.20, 0.18, 0.12. Comparative analysis against traditional machine learning models consistently shows the MDNN’s superior performance by indicating its potential to innovate stress management via personalized music therapy in educational backgrounds.