Research On Modulation Pattern Recognition Of Communication Signal Based On Feature Fusion

Communication signal modulation pattern recognition technology has important research significance in both civil and military fields.The increasing variety of communication signals and complex electromagnetic environments also pose new challenges to existing signal modulation recognition methods.Under low signal-to-noise ratios,the characteristics of signals are greatly affected by noise,and the limited information provided by a single feature data may not be able to characterize certain signals,resulting in low recognition accuracy for certain types of signals or types of signals,thereby affecting the application of signal modulation recognition technology.Choosing suitable data with different signal characteristics and fusing them may provide a promising technical route for signal modulation recognition.In order to solve the problem of low signal recognition rate in the above situations,this paper designs a multi feature fusion algorithm that combines the instantaneous feature sequence of communication signals with deep learning,and designs and compares modulation recognition classifiers based on the sequence characteristics,effectively improving the accuracy of instantaneous feature recognition in low signal-to-noise ratios.The specific research content of this article is as follows:Firstly,this paper systematically introduces the basic principles of communication signals,the theoretical basis of modulation recognition,the extraction algorithm of timefrequency domain features,and the relevant knowledge of convolutional neural networks,laying a theoretical foundation for the preprocessing feature extraction and classifier design in the subsequent research of modulation pattern recognition methods for communication signals based on feature fusion.Secondly,aiming at the characteristics of three types of communication signals,namely,multiple amplitude keying(MASK),multiple frequency shift keying(MFSK),and multiple phase shift keying(MPSK),the instantaneous amplitude and frequency sequence extraction algorithm based on time-frequency analysis and the nonlinear instantaneous phase extraction algorithm are used to extract the instantaneous characteristic sequence of the signal,obtaining the instantaneous amplitude information,frequency information,and phase information of the modulated signal,It provides more significant classification samples for subsequent modulation recognition methods based on multi branch fusion neural networks,improving the recognition accuracy of signals under low signal-to-noise ratio.Finally,in view of the problem that the recognition effect of traditional recognition methods is excessively dependent on artificial feature extraction and the recognition rate is not high under the condition of low signal-to-noise ratio,this paper proposes to use the deep learning technology for multi-feature fusion to improve the classification recognition rate of modulation recognition.Four different neural network classifier models with parallel fusion of three feature branches are designed and built.The modulation recognition simulation experiments of nine communication signals,MASK,MFSK and MPSK,are compared and analyzed.The modulation recognition process based on instantaneous feature fusion under low signal-to-noise ratio is realized through software simulation and hardware measurement.In summary,this article proposes an instantaneous amplitude and frequency sequence extraction algorithm based on time-frequency analysis,a nonlinear instantaneous phase extraction algorithm,and a recognition algorithm that combines deep learning.Starting from using instantaneous feature sequences instead of traditional instantaneous feature statistics and designing an optimized time convolutional neural network classifier structure,the accuracy of modulation recognition is improved.When the signal-to-noise ratio is-4d B,the average recognition rate can reach over 90%.The experimental results show that the proposed method can significantly improve the accuracy of communication signal modulation types under low signal-to-noise ratio.