Modulation Pattern Recognition Based On Multi-feature Fusion

The modulation signal is getting more and more complex in the modern communication system.A variety of modulation signals should be identified.But currently,the algorithms about signal identification lack of applicability.On the other hand,the categories of signal which can be identified are limited.the algorithms just use single feature parameter to identify the signal,definitely it is hard to identify signals clearly if the signals mix a lot of noises.In this paper,due to the lower recognition rate in the environment contains noise,with the feature parameter extraction,improvement and optimization of the classifier,13 kinds of typical digital signal modulation methods,such as MASK,MFSK,MPSK,MQAM,are studied in the lower SNR environment.About the feature parameter extraction method,this paper use a normalized sample entropy feature,which can characterize the complexity of the data.Aiming at the problem that the recognition rate of a single feature parameter is lower,a binary fusion algorithm based on multi-feature has been proposed.A number of characteristic parameters are fused into the binary feature parameters which can be used to identify the modulation signals,and then the pattern recognition is mainly used for the recognition of four kinds of classical modulation types(MASK,MFSK,MPSK,MQAM).In this paper,an AP-FCM joint clustering algorithm based on constellation is proposed,and the clustering validity criterion is used to detect the clustering performance.The adaptive reconstruction and restoration constellation diagram is simulated,so that it can recognize 4PSK,8PSK,16 PSK,16QAM,32 QAM,64QAM.Based on the amplitude spectral density and the sample entropy feature,the MASK modulation signal is identified by the fusion which can identify the 2ASK,4ASK,8ASK modulation signal.The MFSK modulation signal that through the spectrum and the standard deviation characteristic of the peak spacing to identify the 2FSK,4FSK,8FSK modulation signal.Finally,the simulation interface design of AP-FCM clustering algorithm based on MATLAB GUI is designed.In this paper,the simulation and experiment of the improved and optimized algorithm,the algorithm can accurately identify the modulation type and have higher recognition rate when the SNR is low,and use the adaptive reconstructionof AP and FCM algorithm,The constellation diagram is restored and the recognition is high,which proves the validity and correctness of the pattern recognition after multi-feature fusion algorithm.