Research On OFDM Signal Modulation Recognition Method Based On Short-time Fourier Transform And Convolutional Neural Network

Due to the rapid development of communication technology and the increasing complexity of communication equipment,communication signals with a variety of modulation methods have appeared in the electromagnetic environment.As a key work link in radio spectrum monitoring,cognitive radio,and software radio,communication signal modulation mode recognition is a technique for analyzing received signal samples to obtain their modulation mode and other modulation parameters in the absence of prior information and noise interference.Orthogonal Frequency Division Multiplexing(OFDM)technology is a new type multi-carrier modulation technology that is widely used in civil and military fields.In the complex electromagnetic space environment where a variety of digital communication signals and noise coexist,it is more difficult to effectively recognize OFDM signals,and there are few related research results.Therefore,this paper introduces the deep learning method in artificial intelligence technology into the field of OFDM signal modulation recognition,an algorithm using short-time Fourier transform and VGG-16 convolutional neural network is proposed to solve the modulation recognition problem of OFDM signal and single carrier signal under the condition of low signal-to-noise ratio.The main research contents of this article are as follows:1.Introduce the basic theory of digital communication.The basic theories of four single-carrier modulation signals such as MASK,MFSK,MPSK,and MQAM are analyzed.The modulation principle of multi-carrier system is analyzed,and the basic principle and basic structure model of the OFDM modulation system are presented.The basic theory of communication signal cyclostationarity is analyzed.2.Introduce deep learning and convolutional neural network theory.The development process and areas of expertise of deep learning are introduced,the relationship between the three technologies related to deep learning is given,the model,training process and over-fitting problem handling methods of the basic neural network are analyzed,and activation functions and their characteristics are analyzed.The focus is on the relevant theories of convolution neural network(CNN),the basic structure,main characteristics,and related operations of CNN are introduced in turn,the Softmax classifier used in the output layer of CNN is analyzed.3.Propose an OFDM signal modulation recognition algorithm using short-time Fourier transform and VGG-16 convolutional neural network.Firstly,the algorithm uses the cyclostationarity of the communication signal with a symbol interval to perform short-time Fourier transform(STFT)with a window length of the symbol interval on the OFDM signal and several single-carrier signals in each symbol period,and convert them into corresponding time-frequency images,and the length of the window is the symbol interval;then,the time-frequency image is sent to the VGG-16 convolutional neural network model as the signal shallow feature for training,and the network convolution layer and pooling layer automatically extracts the feature information of the OFDM signal and several single-carrier signals;finally,the Softmax classifier is selected in the network output layer to complete the modulation classification of the OFDM signal and several other single-carrier signals.The overall recognition results of the algorithm and the complexity of the algorithm are analyzed in the simulation,and the impact of the algorithm on the performance of the modulation recognition system from the perspective of different convolution kernel sizes,different activation functions,and different optimization algorithms are analyzed.The OFDM signal modulation recognition method based on short-time Fourier transform and convolutional neural network improves the accuracy of OFDM signal modulation recognition under low signal-to-noise ratio,which verifies the effectiveness of this paper.