Heart Sounds Classification Using Short-Time Fourier Transform and Gray Level Difference Method

Abstract

The heart sound coming from the patient is observed using a stethoscope, which is a medical tool to determine the patient's condition. The technique for this observation is called auscultation. This sound describes the condition of a person's heart. Because auscultation relies on the experience and knowledge of doctors, various methods for analyzing heart sounds are automatically developed by researchers. In this study, a method for classifying normal heart sounds and murmurs is proposed using the grey-level difference matrix (GLDM) feature taken from the short-time Fourier transform (STFT) plot. The STFT plot is converted into an image then the GLDM characteristics are calculated as input for the support vector machine as a classification. The experimental result shows that the highest accuracy of 83% is achieved using STFT 200-100 in four directions of GLDM. Even though this accuracy is not as high as the previous research, the proposed method is still open for exploration, such as distance selection in GLDM or other image analysis methods.