Automated Normal and Abnormal Heart Sound Classification Using Hybrid Feature Ensemble

<jats:p>The automated analysis of phonocardiogram (PCG) signals is essential for the early identification of cardiovascular diseases. This book chapter presents new algorithm for classification of pathological heart sounds utilizing a hybrid feature ensemble and Support Vector Machine (SVM) classifier. The proposed methodology employs Infinite Impulse Response Constant-Q Transform (IIR-CQT) spectrograms to extract a comprehensive multi-domain feature set that includes spectral, textural, statistical, and distinctive IIR-CQT characteristics. Feature selection technique is employed using Analysis of Variance. An SVM classifier utilizing Linear, Polynomial, and Radial Basis Function (RBF) kernels is implemented for classification. The proposed hybrid feature approach is evaluated on PhysioNet 2016 dataset. The feature fusion approach achieves improved classification performance with the RBF kernel attaining an accuracy, precision and f-measure of 98.20%, 98.0%, and 98.20% respectively.</jats:p>