The automatic identification of the modulation types has been applied to many fields such as signal identification, interference identification, radio interception, monitoring, etc. The objective of automatic modulation identification is to decide the modulation type and estimate the modulation parameters without any priori knowledge about the signal information content. On the basis of previous research, the paper adopted a statistical pattern recognition framework and designed a hierarchical architecture combined classifier based on MLP artificial neural networks (ANN) with some higher-order statistical parameters selected as identification features. The performance of the algorithm is verified through plenty of computer simulations and the influence of data length for the correct identification probability is investigated. The paper also reseach on the extraction of the instantaneous parameters and estimation of the symbol rate of communication signals, and have a preliminary research on the algorithms for modulation identification based on the methods of wavelet transform and fractal dimension. |