A New Automatic Modulation Recognition Method in Test and Measurement Technology

In this paper, a method is presented based on the cyclic spectral features and the neural network classifiers for performing automatic modulation recognition. The process of the automatic modulation recognition using the method is divided into three basic steps: cyclic spectrum analysis, feature extraction and neural network classifier. Because a new method called as maximum likelihood filter method is used to estimate cyclic spectral density of communication signal, better estimate performance can be acquired in the scene of short length of data. Probability neural network algorithm improves the performance of classifier. Some computer simulation results showing the performance of the method in this paper are improved. The method can efficiently recognize almost all currently used modulation types and the recognition accuracy rate is over 95% at the SNRs of 10 dB.