International Research Journal of Education and Technology

Heart murmurs, caused by turbulent blood flow or structural abnormalities, are critical indicators of cardiovascular diseases such as valvular disorders and congenital heart defects. Conventional auscultation methods rely on the expertise of clinicians and are prone to subjectivity, leading to misdiagnoses. This project leverages deep learning algorithms to accurately classify acoustic heart murmurs from phonocardiogram (PCG) signals. The system employs advanced signal processing techniques to denoise and segment PCG recordings, followed by feature extraction using wavelet transforms and Mel-frequency cepstral coefficients (MFCCs). A hybrid CNN-RNN architecture is developed for murmur classification, ensuring robust handling of temporal and spatial data patterns. The model's performance is evaluated using metrics like accuracy, sensitivity, and specificity. A web- based user interface facilitates real-time analysis, allowing clinicians to upload PCG data and visualize results with diagnostic confidence.

Key Words: Heart murmurs, phonocardiogram (PCG), deep learning, CNN-RNN, MFCC, signal processing, cardiovascular diseases, automated diagnosis.