A multi-level network for radio signal modulation classification
Lei Peng, Wenzhong Qu, Yan Li Zhao, Yunkun Wu · 2019
A new multi-level network combined the Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) is proposed to classify modulation types of signals. The network takes the advantage of CNN in feature extraction and robust to noise and the advantage of LSTM in time sequence analysis. The experiments based on a GNU radio data set shows that the classification accuracy is about 93% for ten different modulation type signals at SNR above 2 dB, which is much better than the conventional CNN. The classification accuracy is about 95% for 8 digital modulation types at SNR above 2dB.