کنفرانس مهندسی برق ایران

صفحه اصلی / بیست و نهمین کنفرانس مهندسی برق ایران

Heart Abnormality Classification by Phonocardiogram Analysis Using Fusion in Feature and Decision Levels

نویسندگان :

Hossein Rahmati¹ Hassan Ghassemian² Maryam Imani³

1- دانشگاه تربیت مدرس 2- دانشگاه تربیت مدرس 3- دانشگاه تربیت مدرس

کلمات کلیدی :

Decision Fusion, Feature Fusion, Dempster-Shafer combination Rule, PCG Signal

چکیده :

Abstract—The audio signal generated by the mechanical activity of the heart provides useful information about the function of the heart valves. The advantage of this method is fast, inexpensive and non-invasive. Due to human auscultator limitation and non-stationary characteristic of phonocardiogram signals (PCG), diagnosis based on the sounds that are heard via a stethoscope is a difficult skill. Therefore, an automatic system to classify biomedical signal PCG, which is recorded by a digital stethoscope, is required. Accurate segmentation of the heart sound signal requires the corresponding ECG signal. But, acquiring ECG is generally expensive and time consuming. This study has proposed a segmentation free system for classification of PCG signals. In order to extract appropriate features of PCG signals, various methods such as, non-uniform filter banks based on maximum entropy, wavelet transform (WT) and powerful features extracted by MFCC that are fused with fractal features are used. Features are given to three classifiers: Support Vector Machine (SVM), K-nearest neighbor (K-NN) and maximum likelihood (ML). The Dempster-Shafer combination Rule (DSR) is utilized in decision fusion step. These experiments were performed on six popular datasets to evaluate the performance of various methods. One data set consists of four classes and the rest consists of two classes (normal and pathological). Sensitivity, Specificity and kappa coefficients were obtained from all six data sets and it is observed that the proposed method is superior to other methods.