Research On End-to-End Modulation Recognition Method Of Wireless Communication Signal

In communication system,signal carrier modulation is the process of converting a signal at the transmitter into a signal suitable for transmission in the transmission medium.Signal modulation classification refers to the technology that determines the modulation mode of the received signal by analyzing the received signal under the condition of unknown signal parameters.Therefore,the recognition of signal modulation mode is the basis to ensure that the information can be transmitted and decoded correctly.However,the recognition of these complex communication modulation modes is increased with the increasing complexity of communication environment,complex communication modulation modes such as MAPSK and MQAM are applied to communication systems.In order to solve the problem of low recognition rate among MPSK,MQAM and MAPSK in complex scenes,this paper analyzes the time-domain characteristics of the signal as well as its spectral and cumulant characteristics.Based on previous studies,three modulation classification algorithms are developed:(1)Modulation recognition research based on feature extraction.Firstly,the constellation diagram of the signal is introduced,and then the spectral characteristics,high order cumulants and M-power spectrum of the signal are expounded.Five characteristic parameters are selected according to the difference of spectrum,cumulant and M power spectrum characteristics,and the specific modulation mode of the characteristic parameters is analyzed,and the threshold of the characteristic parameters is obtained.According to the threshold identification process,the modulation recognition research of multi-characteristic parameters is carried out.The results show that this method has a high recognition rate,but it needs to be manually extracted and selected for specific modulation modes,and cannot be extended to other modulation modes,so it has limitations.(2)Research on modulation recognition based on convolutional neural network.On the basis of convolutional neural network model structure,this paper develops the application of CNN model in modulation mode classification,and proposes an Atten＿CNN model.Firstly,the time domain waveform characteristics of different modulated signals are analyzed and the transmission process of signals is introduced.According to the time domain waveform characteristics of different modulated signals,the data set with the characteristics of the received signals in the actual communication is generated.Then,convolutional neural networks are used to train modulation recognition models on these data sets.The classification and recognition of different modulated signals are carried out based on the modulated signal data set and sequence model.Finally,four models based on convolutional neural network for modulation recognition research are selected as comparison experiments,the five models are trained,the model performance is analyzed,and the experimental results are compared.The overall performance of Atten＿CNN model proposed in this paper is optimal.(3)Research on modulation recognition based on Transformer.Aiming at the problems of limited receptive field and high time complexity of CNN,this paper proposes modulation recognition research based on Transformer model in order to capture the long-distance dependence of data series.The multi-head attention mechanism in Transformer helps models capture long distance dependencies to better extract the characteristics of timing signals.On the basis of this model,an improved CTNN model is proposed by combining the convolution layer with Transformer-Encode,which can improve the learning of local features and the recognition rate of the model through the translation invariance of the convolution layer.