Document Type : Original/Review Paper
Authors
1 Faculty Member of Electrical Engineering Department
2 Electrical Engineering Department, Shahrood University of Technology, Shahrood, Iran
Abstract
A phonocardiogram, which is a graphical representation of the heart sound signal, includes not only the low-frequency components known as fundamental heart sounds but also high-frequency ones called murmurs. The physiological and pathological information contained in heart sounds can be utilized for diagnosing heart diseases. However, due to the transient and non-stationary nature of the heart sound signal, human hearing limitations, and the lower energy of pathological sounds compared to normal sounds, diagnosing heart diseases using these signals is challenging.
The time-frequency domain enables access to the time-varying frequency components of nonstationary signals. In this research, the Stockwell transform is employed to extract time-frequency features. Linear Discriminant Analysis (LDA) is used for feature reduction, while Support Vector Machine (SVM) and k-Nearest Neighbor (kNN) are applied for classification. Furthermore, a publicly available Phonocardiogram (PCG) database consisting of five categories of PCG recordings is utilized to validate the proposed technique. According to the results, the proposed method demonstrates high classification accuracy in binary problems (normal and abnormal) with 99.22% accuracy using either SVM or kNN. For the five-class problem (normal, aortic stenosis, mitral regurgitation, mitral stenosis, mitral valve prolapse), accuracy reaches 95.78% and 96.67% with SVM and kNN classifiers, respectively.
Keywords
Main Subjects
)