Analouge and Digital Modulation Classification Using Deep Neural Networks

Automatic modulation classification is crucial for cognitive radio as well as for military and civilian applications. All signals undergo modulation before transmission; without prior knowledge of the modulation scheme, demodulation of the signal becomes a challenge. In this study, we investigate the classification of modulation schemes using Deep Neural Networks. Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) are currently among the most powerful deep learning algorithms, excelling in this area. The first step of data generation is carried out with extensive care using different random noise samples ranging from −10 dB to 10 dB. In the second step, the modulated signals are then transformed into digital images by applying Continuous Wavelet Transform (CWT), which makes signal analysis easier. At last the transformed images are separated in to training and testing sets with ratio of 70% and 30%. Overall, the results of the model shows a good accuracy of 79%.