Edge-AI Enabled Affordable Wearable for Intelligent Monitoring of Physiological Patterns Associated with Neuro-Cognitive Disorders

Across the globe, mental and neurological diseases including stress, tremors, seizures, and cardiac arrhythmia cause a vast number of people to be afflicted. However, the prevailing principle of early detection is thwarted by periodic clinical examinations and sophisticated expensive equipment, especially in rural or resource-limited areas. The usual mechanisms cannot guarantee a continuous and conveniently accessible trend of monitoring that will lead to interventions without a delay in remedying the prognosis. This study develops an affordable, compact wearable device using an ESP32 microcontroller integrated with sensors like MAX30100 (HR/SpO2), AD8232 (ECG), MPU6050 (motion/tremor), LM35/DHT22 (temperature/humidity), and a capacitance sensor (sweat). A hybrid detection system combines rule-based thresholds with an embedded Random Forest ML model, trained offline on 200 empirical physiological recordings obtained under controlled experimental conditions and subsequently enhanced using statistical data augmentation techniques, such as Gaussian noise and bootstrapping, to augment data variability during model development. Features include HR, HRV (SDNN, RMSSD), ECG R-R intervals, RMS acceleration, and trends. The model supports multi-class classification for conditions like stress and seizures, with on-device alerts via buzzer and OLED. Initial experimental validation with a restricted participant dataset and 5-fold cross-validation exhibited encouraging classification performance under regulated testing settings. The findings underscore the practicality of combining multi-sensor physiological monitoring with integrated machine learning for real-time health status evaluation. Nonetheless, additional extensive and varied clinical investigations are necessary to assess the system’s generalisability across distinct groups. Confusion matrix analysis showed minimal errors, with ECG HR and tremor acceleration as top predictors. The device, miniaturized into a wristband costing under $15, enables offline edge computing with AES-256 encryption. This innovation bridges the gap between consumer wearable technology and intelligent physiological monitoring systems, particularly for resource-limited environments, promoting health equity in underserved areas. By enabling real-time, personalized monitoring, it paves the way for telemedicine, reducing healthcare burdens and fostering proactive interventions for neurological disorders.

Wearable healthcare, random forest classifier, offline machine learning, feature extraction, physiological signal processing, synthetic dataset, statistical data augmentation, scoring system

1. D. Dias and J. P. S. Cunha, “Wearable Health Devices—Vital Sign Monitoring, Systems and Technologies,” Sensors, vol. 18, no. 8, Art. no. 2414, 2018. doi: 10.3390/s18082414. [Google Scholar] [Crossref]

2. S. M. R. Islam, D. Kwak, M. H. Kabir, M. Hossain, and K. S. Kwak, “The Internet of Things for Health Care: A Comprehensive Survey,” IEEE Access, vol. 3, pp. 678–708, 2015. doi: 10.1109/ACCESS.2015.2437951. [Google Scholar] [Crossref]

3. Fitbit Inc., “Fitbit Sense: Advanced Health Smartwatch with ECG, Stress Management and Health Metrics,” Technical Documentation, 2021. [Google Scholar] [Crossref]

4. Y. Zheng et al., “Embedded Deep Learning for Biosignal Processing in Wearable Health Monitoring: Challenges and Opportunities,” Journal of Biomedical Informatics, vol. 107, Art. no. 103451, 2020. doi: 10.1016/j.jbi.2020.103451. [Google Scholar] [Crossref]

5. J. Pan and W. J. Tompkins, “A Real-Time QRS Detection Algorithm,” IEEE Transactions on Biomedical Engineering, vol. BME-32, no. 3, pp. 230–236, 1985. doi: 10.1109/TBME.1985.325532. [Google Scholar] [Crossref]

6. G. Valenza, L. Citi, and R. Barbieri, “Estimation of Instantaneous Sympathetic and Parasympathetic Dynamics From Heartbeat Intervals,” IEEE Transactions on Biomedical Engineering, vol. 61, no. 11, pp. 2828–2838, 2014. [Google Scholar] [Crossref]

7. O. Amft and P. Lukowicz, “From Context Awareness to Socially Aware Computing,” IEEE Pervasive Computing, vol. 9, no. 3, pp. 32–38, 2010. [Google Scholar] [Crossref]

8. F. Sabry, T. Eltaras, W. Labda, K. Alzoubi, and Q. Malluhi, “Machine Learning for Healthcare Wearable Devices: The Big Picture,” Journal of Healthcare Engineering, vol. 2022, Article ID 4653923, 2022. [Google Scholar] [Crossref]

9. G. Vos, K. Trinh, Z. Sarnyai, and M. Rahimi Azghadi, “Generalizable Machine Learning for Stress Monitoring From Wearable Devices: A Systematic Literature Review,” International Journal of Medical Informatics, vol. 178, Art. no. 105295, 2023. [Google Scholar] [Crossref]

10. Y. Lv, W. Dou, S. Liu, and M. Hossain, “Smart Wearable Systems for Health Monitoring,” Sensors, vol. 23, no. 5, Art. no. 2479, 2023. [Google Scholar] [Crossref]

11. Z. Yin, F. Liu, X. Zhang et al., “Leveraging Machine Learning for Personalized Wearable Biomedical Devices: A Review,” Journal of Personalized Medicine, vol. 14, no. 2, Art. no. 203, 2024. [Google Scholar] [Crossref]

12. Yusuf, T. Al Jaber, and N. Gordon, “Comprehensive Health Tracking Through Machine Learning and Wearable Technology,” Journal of Data Science and Intelligent Systems, vol. 5, no. 2, 2025. [Google Scholar] [Crossref]

• What the Desert Fathers Teach Data Scientists: Ancient Ascetic Principles for Ethical Machine-Learning Practice
• Comparative Analysis of Some Machine Learning Algorithms for the Classification of Ransomware
• Comparative Performance Analysis of Some Priority Queue Variants in Dijkstra’s Algorithm
• Transfer Learning in Detecting E-Assessment Malpractice from a Proctored Video Recordings.
• Dual-Modal Detection of Parkinson’s Disease: A Clinical Framework and Deep Learning Approach Using NeuroParkNet