Communication Signal Modulation Mechanism Based on Artificial Feature Engineering Deep Neural Network Modulation Identifier

Based on the characteristics of time domain and frequency recognition theory, a scheme is designed to complete modulation identification communication signals including 16 analog digital modulations, involving 10 different eigenvalues in total. In in-class FSK signal, feature extraction carried out, statistical algorithm spectral peak number proposed. This paper presents method calculate rotation degree constellation image. By calculating modifying clustering radius, rate QAM signal improved significantly. Another commonly used for constellations based Radon transform. Compared with proposed algorithm, has lower computational complexity higher accuracy under certain SNR conditions. discriminator deep neural network, features cumulative are extracted as inputs, modified linear elements neuron activation functions, cross-entropy loss functions. recognitor network cyclic constructed signals. The automatic recognizer implemented CPU GPU, which verifies network. experimental results show that artificial good classification both training set test set.