Heart Sound Classification Using Gaussian Mixture Model
8 Pages Posted: 27 Sep 2018
Date Written: August 1, 2018
Abstract
The heart sound extraction by Gammatone Cepstral Coefficient (GTCC) in combination with the Gaussian Mixture Model (GMM) has been demonstrated for modeling of each heart sound class. In this work, gamma tone cepstral coefficient was used as a feature extraction technique, which employed to recognize the human heart sound disease. To evaluate the system, the trained system was applied on two data set, recorded heart sound from two different input hardware. The system has been tested to automatically diagnose the abnormal heart sounds such as whooshing, rumbling, turbulent fluid caused by an imperfect valve opening. The result show the diagnose accuracy of 32,69% from dataset A with 122 trained data, 23,07% with 72 trained data, and from dataset B of 70.25% with 310 trained data and 39.48% with 93 trained data.
Suggested Citation: Suggested Citation