Modulation Signal Recognition Based On Neural Network

The communication technology develops rapidly at present,the communication environment is becoming more and more complex,in order to ensure that both sides of the communication can delivery information accurately,safely and efficiently in such a complex communication environment,communication signal takes different modulation modes.Communication signal Classification is an important part in the field of signal analysis,its main task is to confirm the modulation mode of received signal under the condition of mutiple signal environment and noises,so as to provide the basis for further analyzing and processing the signal.The main work of this dissertation is as follows:1.The thesis compares 2ASK,4ASK,2FSK,4FSK,BPSK,QPSK six kind of digital modulation signals in both time and frequency domains,and analyses the difference of power spectrum.2.According to the characteristics of the modulation signal,five kinds of characteristic parameters are extracted from its instantaneous amplitude,instantaneous frequency and instantaneous phase,which enhances the distinction by improved the parameters.3.Chapter 3 emphasizes the influence of modulation signal's number of symbols and the oversampling ratio on the recognition rate,and elaborates the rationality that number of symbols is 2048 and oversampling ratio is 8 from the perspective of simulation experiments.4.Chapter 4 elaborates the principle and the topology structure of the BP neural network,the RBF neural network,the PNN neural network and support vector machine(SVM).It is theoretically proved that the neural network can be used in the classification of modulation signal.5.It is successfully completed that classification of the six kinds of signal respectively using method of the traditional decision theory,the BP neural network,the RBF neural network,the PNN neural network and support vector machine(SVM)in the MATLAB platform,and the results are simulated and analysised under the condition that the signal-to-noise ratio are respectively-3dB,5 dB,10 d B,the experimental results show that apart from the traditional decision theory method which requires signal-to-noise ratio is more than 10 d B,three kinds of neural network and support vector machine classifier can comparatively accurate identify the six kinds of modulation signal when the signal-to-noise ratio is greater than 5 dB,that achieves the goal of classification of the modulation signal.Finally this paper presents characteristics and differences of neural network and support vector machine(SVM)used for classification of digital modulation signal,and compared the performance and characteristics of three kinds of neural network(the BP,RBF,PNN),which provides reference both in theory and in practice for the engineering application.