%0 Journal Article
%A LI Shu-fang
%A MA Yue
%A TIAN Hao-min
%A YIN Liang
%T Signal Recognition Based on Signal Power Spectrum Fitting Feature Extraction
%D 2018
%R 10.13190/j.jbupt.2017-183
%J Journal of Beijing University of Posts and Telecommunications
%P 114-118
%V 41
%N 2
%X Aiming at the problem of communication signal service type identification, the author proposes a polynomial fitting factor that extracts the power spectrum of the signal using the linear regression algorithm and the polynomial fitting model in machine learning field as a unified feature of the signal to construct the training set. A multi-layer fully connected neural network classifier model was built on depth learning platform. Compared with the traditional ones, this method has advantages of unifying the radio signal without the need of individual service to extract the personalized features. The power spectrum data of code division multiple access(CDMA) uplink, CDMA downlink, extended global system for mobile(EGSM) uplink, EGSM downlink, wireless local area networks (WLAN) and long time evolution (LTE) signals in the actual radio monitoring data are selected as data set, and 97% classification accuracy is verified. that this method is proven feasible.
%U https://journal.bupt.edu.cn/EN/10.13190/j.jbupt.2017-183