Research On Method Of Modulation Classification Of Digital Communication Signals Based On Compressive Sensing

As core technology of spectrum management and communication countermeasures,modulation classification of non-cooperative wireless communication signals plays an important role in civil and military communication areas.In most sceneries,monitoring objectives have the characteristic of owing wide spectrum,which leads to the serious challenge of high sampling rate based on Nyquist sampling theory.Moreover,high sampling rate can also lay great pressure on signal processing ability and saving space.Aiming at the challenge raised above,the paper carries out the research of wireless signals with wide spectrum from theory,signal processing methods,as well as experimental verification,based on the compressive sensing theory proposed by Candes and Donoho in 2006.In this paper,we analyze the sparsity characteristic of signals with wide spectrum,construct spectrum and statistical features as identification features,recover the constructed identification features using compressive samples,and then carry out researches on signal modulation classification theories.Then,the paper show simulation verification results for the proposed method,proving that our method can lower the sampling rate for signal modulation classification,improve the signal processing speed and save the storage space effectively and obviously.The main contributions and innovation points are listed below.The paper proposes a feature-reconstruction method based on high-order moments and spectrum features,which applies digital signal modulation identification methods based on features.The method analyze the sparsity and difference of signals as basis of feature selection,and then determine the high-order moment and signal spectrum feature as identification characteristics.The paper construct the linear reflection relationship between signal compressive samples and identification features,which combines modulation classification technology and compressive sensing theory effectively and lays the foundation of construction for signal features using compressive samples.What’s more,on basis of recovery of signal identification features,the paper propose the theoretical method on digital signal modulation classification using compressive samples systematically,accompanied by simulation verification.The results show that for different modulation modes,the correct classification probabilities are also different.For most signals modulated by MPSK,MFSK,however,the classification can be accurate in relative low SNR(Signal-to-noise ratio).For MQAM,the reason for low identification probability is analyzed and efficient solutions are proposed,which can improve the performance of this method significantly.