Multilevel wavelet packet entropy: A new strategy for lung sound feature extraction based on wavelet entropy

Wavelet packet entropy (WPE) is one of the entropy measurement methods based on wavelet transform. If wavelet entropy (WE) is used in discrete wavelet transform (DWT), then WPE used wavelet packet decomposition (WPD) for entropy calculation. Various entropy measurement techniques are used in WPE calculations and generate 2
^N
parameters where N refers to the decomposition level. In this paper, we proposed a new method based on WPE called multilevel wavelet packet entropy (MWPE). With the proposed method, the number of parameters produced is N, the wavelet decomposition level. The accuracy up to 97.98% was obtained for five classes of lung sound. The proposed method yielded accuracy higher than the use of one decomposition level in WPE. This method can also be used for the features extraction of the biological signals such as electrocardiogram (ECG), or electroencephalogram (EEG).