Low-SNR Modulation Recognition based on Deep Learning on Software Defined Radio

Automatic modulation classification (AMC) and recognition (AMR) of received wireless signals have a significant role for various commercial and military areas. These methods are able to identify the modulation type and recognize the received signal by extracting discriminating features from the signals. Deep neural network (DNN) offer a great tool that assist the identification of signal modulation because of its capability to extract complex features from the received signals. In this work, we propose a convolutional network model to classify the modulation type of a wireless signal at low-SNR values. The experimental results demonstrate that the proposed model correctly classify 72% digital signals at -4 dB. The accuracy can be increased if the similarities between QAM4 and QAM64, 8PSK and QPSK is reduced.