Automatic Auscultation Classification of Abnormal Lung Sounds in Critical Patients Through Deep Learning Models

摘要

This research aims to use the output signals of a stethoscope and classify them through deep learning models automatically. In this research, the dataset consists of four classes, normal, wheezing, crackles, and unknown are used. To effectively classify each signal, we use the spectrogram generated by the short-time fast Fourier transform as the feature value of each lung sound signal and found the best parameters to do model selection. Besides, we also adopt Depthwise separable (DS) convolution technic, and refer to the architecture of Mobile-Net, to achieve the purpose of high accuracy and low model parameters.