ML-Based Analysis of Phonocardiogram Signals for Heart Sound Classification

DOI

10.2991/978-94-6239-713-2_14How to use a DOI?

Keywords

Cardiovascular; PCG; Machine Learning; Accuracy: Heart Sound

Abstract

Cardiovascular diseases are one of the major components of the global health budget. Phonocardiography (PCG), the recording of heart sounds, offers a non-invasive method for cardiac assessment, with potential for automated analysis to aid diagnosis. This study compares the effectiveness of five distinct machine learning approaches for classifying PCG recordings as ‘normal’ or ‘abnormal.‘ Data were collected from 40 participants (ages 10–30) using a PCG machine at four standard auscultation locations. Recordings were classified by five experts. Five models were implemented and evaluated: two Random Forest (RF) classifiers using aggregated Mel-Frequency Cepstral Coefficients (MFCCs) from different cardiac segments, a 1D Convolutional Neural Network (CNN) using sequences of cycle-based features, a Bidirectional Long Short-Term Memory (BiLSTM) network using similar cycle sequences augmented with timing features, and a Transfer Learning approach using DenseNet201 on Mel Spectrogram images. Performance was evaluated using Test Accuracy and ROC AUC. Results showed that the RF models performed poorly, particularly in class discrimination (Test ROC AUC < 0.4). The deep learning models significantly outperformed RF, with the BiLSTM (Test Acc: 0.7453, Test ROC AUC: 0.8046) and DenseNet201 (Test Acc: 0.749, Test ROC AUC: 0.79) achieving the highest test scores. However, the BiLSTM demonstrated better alignment between validation and test performance compared to the DenseNet model, which showed signs of potential overfitting or poor generalization based on lower validation scores. The findings highlight the importance of sequence modeling and appropriate feature representation for PCG classification and suggest the BiLSTM approach offered the most robust performance in this evaluation.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - S. V. H. Nagendra
AU  - Vijay K. Pandey
PY  - 2026
DA  - 2026/06/25
TI  - ML-Based Analysis of Phonocardiogram Signals for Heart Sound Classification
BT  - Proceedings of the International Conference on Advances in Computing Technology and Artificial Intelligence (COMPUTATIA 2026)
PB  - Atlantis Press
SP  - 192
EP  - 206
SN  - 2589-4919
UR  - https://doi.org/10.2991/978-94-6239-713-2_14
DO  - 10.2991/978-94-6239-713-2_14
ID  - Nagendra2026
ER  -