Comparison of multi-distance signal level difference Hjorth descriptor and its variations for lung sound classifications

Abstract

A biological signal has the multi-scale and signals complexity properties. Many studies have used the signal complexity calculation methods and multi-scale analysis to analyze the biological signal, such as lung sound. Signal complexity methods used in the biological signal analysis include entropy, fractal analysis, and Hjorth descriptor. Meanwhile, the commonly used multi-scale methods include wavelet analysis, coarse-grained procedure, and empirical mode decomposition (EMD). One of the multi-scale methods in the biological signal analysis is the multi-distance signal level difference (MSLD), which calculates a difference between two signal samples at a specific distance. In previous studies, MSLD was combined with Hjorth descriptor for lung sound classification. MSLD has the potential to be developed by modifying the fundamental equation of MSLD. This study presents the comparison of MSLD and its variations combined with Hjorth descriptor for lung sound classification. The results showed that MSLD and its variations had the highest accuracy of 98.99% for five lung sound data classes. The results of this study provided several alternatives for multi-scale signal complexity analysis method for biological signals.