Heart Disease Detection using Audio Dataset

Heart disease detection plays a crucial role in the early diagnosis and management of cardiovascular diseases. In this study, we explore the application of audio-based methods to detect heart diseases using an audio dataset consisting of heart sounds. The dataset, sourced from Kaggle, contains labeled heart sound recordings categorized as either "Healthy" or "Unhealthy." The proposed approach leverages Convolutional Neural Networks (CNN) and Recurrent Convolutional Neural Networks (RCNN) for classification tasks.

For feature extraction, we utilize Mel-frequency cepstral coefficients (MFCC), a widely used technique for capturing the spectral characteristics of heart sounds. Additionally, we enhance the dataset by introducing noise augmentation, pitch shifting, Zero-Crossing Rate (ZCR), and Root Mean Square Error (RMSE) to improve model robustness and generalization. These techniques are effective for handling variations in heart sound recordings due to environmental factors or physiological changes. The CNN and RCNN models are then trained to classify heart sounds into "Healthy" and "Unhealthy" categories.

By processing the audio signals and extracting these key features, our model achieves high classification accuracy, demonstrating the potential of audio-based methods for heart disease detection. This non-invasive and efficient approach can serve as a valuable tool in early-stage heart disease diagnosis, potentially reducing the need for more complex and costly diagnostic procedures.

Keywords: Heart disease detection, audio dataset, Kaggle, heart sound classification, CNN, RCNN, MFCC, noise augmentation, pitch shifting, ZCR, RMSE, feature extraction, deep learning, machine learning, cardiovascular disease, audio-based diagnosis.