Synthesis and Classification of Heart Sounds Using Multi-component Oscillatory Model

Abstract

Analysis of heart sounds (HSs) plays a vital role in the early detection and diagnosis of cardiovascular diseases. In this paper, we propose a multi-component oscillatory model for the representation of both normal and pathological heart sound segments. A half-period sine wave is fitted between every two consecutive zero-crossing points to extract parameters for the proposed model. The segment-representation gets improved with the recursive use of multiple half-wave oscillations. The proposed method is tested and validated with the Computing in Cardiology Challenge (CinC) 2016 database, available publicly on the Physionet archive. The efficiency of the model is demonstrated for the synthesis of HS segments. The performance results of synthesis show that the multi-component oscillatory model provides a highly accurate approximation of the original HS segments. Further, the model parameters are employed for the classification of normal and abnormal HS segments. The proposed method achieves a better performance using support vector machine classifier with RBF kernel.