Automatic Classification of Respiratory Sounds Based on Time-Frequency Analysis and Convolutional Neural Network

Abstract

Auscultation of respiratory sounds is very important for discovery of respiratory disease. However, there is no quantitative evaluation method for the diagnosis of respiratory sounds. It is necessary to develop a system to support the diagnosis of respiratory sounds. In this paper, we propose an algorithm for the automatic classification of respiratory sounds as normal, continuous sound or crackle. Our approach consists of two major components: 1) transformation of one-dimensional signals into two-dimensional time-frequency representation images using the short-time Fourier transform and the continuous wavelet transform and 2) classification of the images using convolutional neural networks. We applied the proposal method to 22 respiratory sound data. As a result, we achieved the accuracy of 79.44 [%] and the area under the curve based on receiver operating characteristic curve of 0.942.