Feature Extraction and Classification Methods for Lung Sounds

Abstract

The lung sounds is a non-stationary signal. It is a major challenge to analyze and differentiate the type of pulmonary disorder based on lung sounds. This paper presents a detailed review of existing methods of feature extraction and classification of Lung sounds for diagnosing the various types of pulmonary disorder. The different methods like spectral analysis, Cepstrum and Mel- Cepstrum, Hilbert Huang Transform, Spectrogram and 2D representation, Wavelet method, time expanded waveform analysis, Hidden Markov model, Auto Regressive model, and Neural Network are being discussed here. All the discussed methods automatically recognise the different types of lung sounds and pulmonary disorder based on features extracted from recorded lung sounds. The paper covered all the suited existing methods which can effectively detect the lung diseases. As per the result of this analysis, it has been found that still more work is required to be done in the screening and classification of chronic Lung diseases. Chronic lung diseases, having similar symptoms and which are very hard to be distinguished and classified. So, therefore, some suitable work needed to be done so that it could effectively support the physicians for taking diagnosis decisions and for giving the correct treatment without any delay in such chronic diseases also.