Find a project you can actually build
Working IoT, AI/ML and embedded builds with abstracts, component lists, documentation, and developer support — ready for your final-year submission.
A Non-Invasive IoT-ML System for Blood Glucose Estimation using AS7341 Visible Light Sensor and ESP32
Accurate and non-invasive blood glucose monitoring remains a critical challenge in biomedical engineering and digital healthcare. Traditional invasive glucose measurement methods, while reliable, are often uncomfortable and unsuitable for continuous monitoring. In this work, we present a novel approach to estimate blood glucose levels non-invasively using the DFRobot Gravity AS7341, an 11-channel visible light spectral sensor, in combination with Machine Learning (ML) and IoT-based real-time visualization. The system employs the AS7341 sensor to capture light absorption and reflectance characteristics from the human skin under controlled lighting conditions. These spectral readings, influenced by the glucose concentration in subcutaneous tissues, are processed and transmitted via ESP32 over Wi-Fi to a Flask-based backend for real-time data storage, model inference, and dashboard display. An ML model (Random Forest, SVM, or Neural Network) is trained on labeled spectral datasets with corresponding finger-prick glucose measurements. The model predicts glucose concentration based on key spectral features extracted from the AS7341 sensor. Real-time results, trend graphs, and alerts are presented on a lightweight Flask web interface accessible from any device. The system prototype demonstrates the feasibility of visible spectrum analysis for glucose trend estimation, with potential for improvement through the integration of NIR sensors and personalized ML calibration. This work contributes toward the development of a portable, cost-effective, and user-friendly non-invasive glucose monitoring solution.