Multi-scale Grey-Level Difference for Lung Sound Classification

Abstract

Lung sounds have information to seek abnormalities in the lung. With digital signal processing, the information in the lung sounds is extracted as the features in lung sound classification. In this paper, texture analysis was used to measure the complexity of lung sound as a feature in lung sound classification. Grey-Level Difference (GLD) method was performed on lung sounds with a number of different scales. Multi-scale GLD has produced accuracy up to 90.12% for five classes of data. Further, gradient entropy individually provided the highest accuracy up to 91.36% for the distance D = 20 and a scale of 1-10.