Feature Extraction and Area Identification of Wireless Channel in Mobile Communication

The rapid development of mobile communication industry has exerted great influence on human life and social development. The uniqueness of mobile communication comes from wireless channels. The wireless channel may have some different feature in different scenarios or regions. It is a hot topic to analyze and extract these characteristics. The measured wireless channel data for three different scenarios are analyzed in this paper. Firstly, the influence of noise and filter on the measurement signal is analyzed. Secondly, the characteristics of envelope statistics, autocorrelation function, multipath intensity distribution function, Doppler power spectrum and time interval correlation function of wireless channel are studied and the new parameters are defined according to the filter characteristics. The differences of these parameters in different scenes are studied, and the required “fingerprint” features are extracted. In this paper, SVM is the basic unit of classifier to solve the problem of recognition and clustering of wireless channel scenes. Using the channel “fingerprint” feature extracted from different scenes to train the SVM model, and using bayesian posterior probability as the criterion, the recognition of the scene can be realized accurately. The adjacent segment clustering algorithm based on SVM can classify the channel paragraphs after segmentation