Discrimination of respiratory sounds based on a new adaptive function

In this paper, we address the problem of discriminating normal breath and adventitious, continuous (e.g. wheeze) sounds based on a new adaptive function. The proposed function is based on fast Gabor time-frequency distribution and autoregressive (AR) averaging. Using a time-frequency (TF) representation of the input signal together with a gain procedure based on AR averaging provides an effective tool to differentiate normal/adventitious respiratory sounds. Various features/measures, using tracheal respiratory sounds, are presented to demonstrate the efficiency of the proposed approach. The presented method is characterized by its simplicity and ease of implementation.