Research On Automatic Modulation Recognition Technology Of Composite Signal Based On SVM And Residual Network

Compound modulation signal is widely used in TT & C system.Because it has the characteristics of low complexity of analog modulation equipment and strong confidentiality of digital modulation,its application in satellite communication,measurement and control,ranging and broadcasting system is growing day by day,and plays an important role in satellite communication.The research on modulation recognition method for complex modulation signal can effectively promote the development of measurement and control technology in modern complex communication environment.At present,the related research of compound modulation signal recognition mainly relies on the recognition algorithm of extracting signal characteristics artificially.The recognition performance of traditional methods is highly dependent on the accuracy of artificially set features,and the generalization ability is poor,and the recognition rate is low under low SNR.Therefore,based on the digital receiver,this paper proposes a composite signal modulation recognition method based on cascaded SVM and residual network for two signal sets with different parameters.Aiming at the fixed bandwidth compound modulation signal set(MTONE-FM?2ASK-FM?BPSK-FM?QPSK-FM)in TT & C system,a modulation recognition algorithm based on all digital receiver and support vector machine(SVM)is proposed.The internal modulation information of FM composite signal is extracted by phase discrimination error of all digital receiver,and the overall graphic features of power spectrum and square spectrum of internal modulation signal are classified and recognized by cascade SVM classifier.Theoretical analysis and simulation results show that,compared with the traditional recognition algorithm based on local signal features and decision tree classification,the recognition accuracy probability of the proposed method is increased by more than 5% for fixed bandwidth composite modulation signal set.For the variable bandwidth composite modulation signal set in the measurement and control system(2FSK-FM signal with the bandwidth of 2 k Hz and 32 k Hz respectively,BPSK-FM signal with the bandwidth of 2 k Hz and 128 k Hz,and 2 k Hz and 128 k Hz respectively)There are differences in the frequency domain images of the above signals,but the differences are small.Therefore,this paper proposes a modulation recognition method of composite signals based on the frequency domain characteristics of internal modulation signals and residual network structure.Extract the internal modulation signal based on the all-digital receiver to obtain the internal modulation frequency domain image,and the image is moved and trimmed to highlight the difference part.Finally,using the characteristics of fast convergence speed and high classification accuracy of the deep residual network classifier,the recognition of different modulation modes and different parameter signals is realized.The residual network model can extract the deeper features of input data,and there will be no gradient disappearance or gradient explosion under the deep network,so it can be used to distinguish images with similar features.The experimental results show that,compared with the widely used constellation and CNN recognition schemes,the recognition performance of the proposed method is improved by more than 3d B on average for the bandwidth variable composite modulation signal set(2FSK-FM,BPSK-FM,QPSK-FM).