Automatic Modulation Recognition Based on Multi-Dimensional Feature Extraction

As an intermediate step between signal detection and demodulation, automatic modulation recognition (AMR) is commonly used in cognitive radio networks to identify different types of communication modulation. A new automatic modulation scheme is proposed, based on decision tree theory, which is a general method for different types of band-limited Gaussian noise modulation types. In particular, by combining the instantaneous statistic feature and high-order cumulants feature, the key features are extracted to realize the blind recognition of analog and digital signals. In addition, a new characteristic parameter AT is proposed to improve the performance of modulation recognition under low signal-to-noise ratio (SNR). The simulation results show that, for all the analyzed signals, when the SNR reaches 3dB, the recognition success rate can reach more than 95%, reflecting the superiority of the method proposed in this paper.