Modification of Grey Level Difference Matrix (GLDM) for Lung Sound Classification

Texture analysis is one of the methods to see the pixel variety in an image. Texture analysis can be done directly on image pixel value or done using transformation. Texture analysis can be utilized on the 1D signal to observe the variation of signal data samples. In this research, texture analysis using GLDM was modified as feature extraction method for lung sound classification. The features were classified using multilayer perceptron (MLP) and support vector machine (SVM) for performance evaluation. The result showed that modified GLDM with distance d = 10 achieved the highest accuracy of 94.9% using five GLDM's features, cubic SVM, and three-fold cross-validation. The result was achieved for five classes of lung sound consist of 99 data. The proposed indicated that texture analysis could be utilized for biological signal analysis, especially respiration sound.