کنفرانس مهندسی برق ایران

صفحه اصلی / سی و دومین کنفرانس بین المللی مهندسی برق

Diagnosis of Heart Diseases based on Processing Heart Sound using Machine Learning

نویسندگان :

Maryam Moulaverdi¹ Akbar Ranjbar²

1- دانشگاه شاهد تهران 2- دانشگاه شاهد تهران

کلمات کلیدی :

Deep learning،Feature extraction،Heart disease diagnosis،Signal processing

چکیده :

In this research, the data related to heart sounds are classified into five general categories. These five categories include normal sound (N), mitral stenosis (MS), mitral regurgitation (MR), aortic stenosis (AS), and mitral valve prolapse (MVP). The important goals of this research were to use an effective method to select better and more efficient features and advanced machine learning techniques to design a pattern and implement a classifier to separate all five categories of heart sounds. For this purpose, features have been extracted using MFCCs, DWT, and Mel Spectrogram methods. Then, by using SVM and RF methods, the results are checked, and a hybrid DNN network is presented using Co-Shuffle-LSTM. Finally, to check the generalization power of the best model, the multi-category cross-validation method has been used. In this method, FCSL and FCSLWM patterns were used, which reached 98.70% and 98.82% correctness in identifying and distinguishing cardiac signals.