Pulmonary Crackle Detection Using the Hilbert Energy Envelope

This paper presents a method for automatic pulmonary crackle detection based on the Hilbert energy envelope (HEE). Automatic detection of crackles in lung sounds offers a non-invasive way of monitoring or diagnosing cardiopulmonary diseases. The algorithm is divided into four main steps: (a) pre-processing, (b) estimation of HEE, (c) thresholding, and (d) applying time width conditions based on crackle two-cycle deflection and initial deflection width. Its performance is tested using a publicly available lung sound dataset of fine and coarse crackles and evaluated by the sensitivity (95.7%), positive predictive value (89.5%), and F-score (91.7%) for crackle detection. The good detection performance indicates the potential of the HEE-based algorithm as an automatic method for crackle detection in lung sound recordings.