This project presents a real-time embedded AI system for heart sound (PCG) classification using a digital stethoscope for early cardiac screening. A localized dataset collected from rural populations is used to train a Support Vector Machine (SVM) model after signal preprocessing, achieving approximately 80% accuracy for classification of normal and abnormal heart sounds. The trained model is independently deployed on a dual-platform architecture consisting of a RISC-V– based embedded system and a mobile application. Both systems support real-time signal processing, visualization, and result display, enabling standalone cardiac screening. This solution enhances accessibility, portability, and diagnostic efficiency, providing a low-cost and flexible approach for early detection of cardiovascular diseases in resource-limited environments.
Tools: Matlab,Python,Debian Linux,Android Studio,Support Vector Machine (SVM),MFCC Feature Extraction.,Butterworth Low-Pass Filter,FFT Analysis
Department: Department of Electrical Engineering
Poster
