Study On Cascaded Modulation Recognition Method Based On Different Features

Modulation identification is the key technology of software radio and noncooperative communication.It occupies an important position in both civil and military aspects.It provides an important basis for subsequent demodulation and other processing.The existing recognition method has a low recognition rate under the condition of low signal-to-noise ratio.How to improve the modulation recognition rate under the condition of low signal-to-noise ratio is still a problem worth studying.In response to this problem,this article conducted the following research:Firstly,the influence of signal-to-noise ratio on the instantaneous characteristics,fractal characteristics and high-order cumulant characteristics of ten kinds of signals to be identified is studied and analyzed,and the instantaneous characteristics and fractal characteristics are improved to make them relatively insensitive to noise.The modulation signals to be identified are {2ASK,4ASK,8ASK,2FSK,4FSK,8FSK,2PSK,4PSK,8PSK,16QAM}.The improved instantaneous characteristics have improved the correct recognition rate of the modulation mode at low noise to a certain extent,and improved the difficulty of recognition between 4FSK and 8FSK and between 4PSK and 8PSK.The improved fractal feature combination has a decisive recognition rate of 6.5% and a three-category recognition rate of 1.012%,and the improved feature combination performs three-category recognition of the recognition signal at-3d B.All reached more than 95%.Secondly,the effect of signal length on the fractal characteristics and higher-order cumulants was studied and analyzed again.The minimum length of the signal that did not affect the recognition results was determined according to the analysis results.At the same time,the influence of the sampling frequency on the fractal characteristics was studied to determine the best recognition effect Sampling frequency.Thirdly,based on feature analysis,a decision tree and a support vector machine recognition model were established to further analyze the effect of the feature on the modulated signal.Research and analysis show that the instantaneous feature is better than the other two features in the case of a limited length.Correct recognition rate,high-order cumulant has better recognition rate for MPSK mode,fractal feature has better recognition rate for large categories,and fractal feature has better recognition rate for MFSK.Finally,according to the research on different features,decision tree and support vector machine to identify the modulation mode,the identification model of serial and parallel modulation based on different characteristics is proposed.The simulation results show that the newly established recognition model improves the recognition rate of the modulation method under low signal-to-noise ratio.