Research Of Digital Modulation Recognition Based On Artificial Neural Networks

Digital modulation recognition has found important applications both in military and civil, while the feature extraction is the preliminary step for digital modulation recognition. This paper proposes two novel features, and combine them and other existing parameters to construct a new feature set based on statistic pattern recognition. This set of features needs no a priori information about the nature of the signal. Later on, we adopt a layered neural network classifier to complete the recognition process. The modulation types classified by this recognizer are 2PSK, QPSK, 8PSK, 2FSK, 4FSK, 8FSK, 2ASK, 4ASK and 16QAM. Simulation results show that the average rate of accurate classification is over 97% at a signal-to-noise (SNR) of 5dB, and more, the system is robust.