A Semi-automatic Heart Sounds Identification Model and Its Implementation in Internet of Things Devices

Abstract

Identification of heart sound signals in the form of a phonocardiogram (PCG) has recently attracted the attention of many researchers along with the development of small devices and global Internet connection in a way to offer automatic illness detection and monitoring. In this work, we propose a semi-automatic envelope-based heart sounds identification method called the Largest Interval Heart Sounds Detection (LiHSD) that exploits the superiority of the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and the cubic spline interpolation to discover several heart sounds' components such as period and location of S1 and S2, an interval of a cardiac cycle, and to obtain the duration and location of murmurs. Evaluation of the proposed system over several life sample data showed promising results comparable to the previous models. The algorithm was able to capture the largest interval of S1 and S2. The examination for normal heart sounds exhibited detection accuracy 98 percent, whereas for anomaly heart sounds samples the detection accuracy ranging from 89 percent to 97.5 percent. Furthermore, the proposed system has been successfully implemented in a real Internet of Things device while eyeing its contribution to the future of the smart healthcare system.