Analysis of Audio
Signals and Hyper-Parameter Optimization of Convolutional Neural Network for Heartbeat
Anomaly Detection
Gandotra Ekta, Gupta Deepak, Mahajan Ria, Sharma Parul and Kumar Harish
Res. J. Biotech.; Vol. 20(12); 255-264;
doi: https://doi.org/10.25303/2012rjbt2550264; (2025)
Abstract
Cardiovascular diseases hold significant importance as global health concerns due
to their alarmingly high mortality rates. These diseases are characterized by the
heart's inability to supply sufficient blood to the body's other organs. This inability
of the heart can have severe repercussions on a patient's health. Cardiovascular
diseases encompass conditions such as coronary artery disease, heart failure, stroke,
hypertension and various other disorders that affect the heart and blood vessels.
A timely and accurate diagnosis is essential for a patient's survival as well as
for averting further loss.
This study presents an efficient method that helps in identifying irregular heartbeats.
The proposed method uses an audio signal dataset that has been compiled from general
public as well as clinical studies. It uses time-frequency heatmaps and deep convolutional
neural network to automate the classification of heartbeat audio signals. An extensive
process of hyperparameter tuning to optimize the learning rate, batch size and the
number of epochs is used to enhance the performance of the model. The experimental
results of the study show the effectiveness of the proposed model in identifying
irregular heartbeats. After doing hyperparameter tuning, the proposed model obtains
an accuracy of 96% on the validation dataset.
