Research On Modulation Recognition Technology Based On Neural Network

Modulation recognition technology of communication signal has always been a very important research area in the field of wireless communication system.The modulation mode of the transmitter is unknown to the receiver in many cases,so it is necessary to identify the modulation mode of the signal at the receiving end.Therefore,modulation recognition of communication has been a very important research content.However,with the rapid development of wireless communication technology,the channel environment of wireless communication becomes more and more complex.When the signalto-noise ratio is low,the received signal is easily distuebed by noisy noise after passing through the channel,which leads to the the power of the signal is small and the effective features extracted when processing the received signal are few,and makes the recognition algorithm can not correctly identify the modulation type of the signal.Therefore,the main purpose of this paper is to further study the modulation recognition technology in digital communication systems under low SNR,aiming at making more accurate recognition of modulation mode in the case of least prior information.The research contents of this paper are as follows:(1)The common digital modulation methods are described in detail,and 10 modulation modes,such as MASK MFSK MPSK and 16 QAM,are selected as the recognition objects.Then we analyse the key technologies of neural network classifier.(2)The features of communication signal including time domain features,continuous wavelet transform features,the high order cumulants features and frequency domain features are extracted on the basis of the existing feature parameter extraction.The simulation analysis of these characteristic parameters is carried out and their classification ability is compared.according to the value of the zero center normalized statistical variance,the features which are less affected by noise are selected as the training features of the classifier.(3)we analyse the training process of the traditional BP neural network theoretically and reseach the classification performance of BP neural network under the selected training characteristics,simulate the recognition rate of modulation recognition based on BP neural network,when the SNR is 5 d B,the modulation recognition rate reaches 70%.(4)we study and analyse the training process and fast learning algorithm of DBN network and proposed a modulation recognition algorithm based on DBN network.We similuate and analyse the recognition rate of modulation mode based on DBN neural network.The simulation results show that the performance of modulation recognition algorithm based on DBN network is significantly higher than the modulation recognition algorithm based on BP neural network in the case of low signal-to-noise ratio.The recognition rate of modulation based on DBN network has reached 90% When the SNR is 3 dB.