Automatic Identification of Abnormal Lung Sounds Using Time-Frequency Analysis and Convolutional Neural Network

This research focuses on the development of a method utilizing signal processing and machine learning techniques to identify abnormal lung sounds, specifically adventitious lung sounds, for diagnosis and monitoring. The proposed algorithm combines short-time Fourier transform (STFT) with convolutional neural networks (CNN) to automatically analyze breath sounds captured by a stethoscope. By employing a band pass filter, noise is effectively removed, facilitating accurate identification of lung sounds. The algorithm classifies abnormal lung sounds, such as crackles and wheezes, with an impressive accuracy rate of 85.27%. This research not only enhances the efficiency of physical examinations but also enables the recording and analysis of lung sounds, thereby offering valuable insights into the progression of treatments. Furthermore, the development of this medical device has significant implications for advancing human healthcare and information retrieval in the field of respiratory medicine.