Feature extraction method for wireless network devices using multivariate empirical mode decomposition

Abstract

Due to noise interference and signal aliasing, the existing methods are difficult to effectively extract the characteristics of the individual characterization of the radiation source, and achieve high-precision identification of the radiation source. So we propose a method which is based on the Multivariate Empirical Mode Decomposition (MEMD) technique, and it can be used to separate wireless frame stray signals, then identify individual devices. Firstly, MEMD method is used to extract the stray signal from the wireless frame signal; then the subtle features in the wireless frame stray signal can be extracted by a series of feature extraction methods; finally, many wireless network device individuals of the same type are classified and identified by a Support Vector Machine (SVM) classifier. The experimental results show that in the identification of five wireless network devices of the same type, the MEMD method is used to extract the stray features and obtain an average recognition rate of 94.59%. At the same time, the recognition rate of the MEMD method is significantly higher than the EMD method.