TiO ₂ /PSi Heterostructure-Based Multiwavelength Photodetector System for AIoT Applications

This work presents the development and testing of an IoT-integrated TiO2/porous-silicon heterostructure photodetector designed for multi-wavelength sensing, targeting applications such as photoplethysmography and bionic eye systems. The sensor is designed to respond to various light sources and accurately detect the corresponding wavelengths. Additionally, the system is capable of recognizing and reconstructing handwritten digits. The in-house developed TiO2/porous-silicon heterostructure sensor was exposed to light from five sources: ambient light, flashlight, red LED, blue LED, and green LED. The flashlight produced the highest response, indicating enhanced sensitivity due to its broader spectrum. Principal Component Analysis (PCA) was employed to effectively segregate the sensorfs response to different wavelengths, improving signal classification accuracy. The sensor data was uploaded to the cloud using the Thingspeak platform for real-time monitoring and IoT-based applications. Machine learning models, such as Random forest, Support Vector Machines (SVM) and K-Nearest Neighbor (KNN) were applied to predict the wavelength, achieving an accuracy of approximately 98% with a low mean squared error (for KNN). Furthermore, flashlight-derived sensor data was mapped onto MNIST handwritten digits, and a denoising autoencoder successfully reconstructed the original digits, demonstrating the systemfs ability to process and analyze complex signals for real-world pattern recognition. Integrating multi-wavelength photodetector with IoT and AI enables accurate, real-time sensing for biomedical and optical applications, paving the way for advanced non-invasive diagnostic technologies.