An improved EEMD model for feature extraction and classification of gunshot in public places

摘要

Ensemble empirical mode decomposition (EEMD) is a noise-assisted adaptive data analysis method. The key of EEMD is to add Gauss white noise into the signal to overcome mode-mixing problem caused by original empirical mode decomposition (EMD). Because the noise in public places is natural noise with alpha stable distribution, in this paper we proposes an improved EEMD by using symmetric alpha stable (SaS) distribution instead of the Gauss distribution, and applies the improved EEMD for extracting gunshot feature. Using the improved EEMD, firstly we decompose gunshot signals into a finite number of intrinsic mode functions (IMF). Then, we use the energy ratio of each IMF components to original signal as gunshot feature for classification. The results of simulating experiment show that the improved EEMD method has good generalization abilities for the feature extraction of gunshot in public noise places.