Detecting Lung Infections with Empirical Mode Decomposition and Neural Networks

This research study is dedicated to leveraging lung sound analysis as a diagnostic tool to differentiate between individuals with pneumonia and those in good health. The preprocessing phase involves critical steps such as segmenting lung sound signals and resampling. The data obtained from the time, frequency, and cepstral domains are effectively harnessed to extract meaningful features. Furthermore, feature selection techniques are employed to identify the most relevant features. The application of a Neural Network classifier yields an impressive accuracy rate of 99.3%. This study underscores the potential of utilizing the Neural Network classifier in conjunction with time and cepstral domain data for the reliable and accurate diagnosis of both pneumonia and healthy individuals.