An optimized Radio Modulation Classifier Using Deep Neural Network

Abstract

Automatic Modulation Classification (AMC) is vastly used in civilian and military equipment for instance; the signals in satellite, GSM, Wi-Fi…etc. The identification and handling of the features of these signals became more sophisticated due to the predominating progression of modern communication technology mainly in the past decade. In the last few years, researchers have begun to study the use of Deep Neural Network (DNN) to identify different types of modulation, and by using it the task has become easier, and we can get a significant improvement in the classification performance compared to many traditional methods. In this paper an automatic modulation Classification model is proposed where deep learning is used to classify different types of modulation at different signal to noise ratios (SNRs), where we optimize the conventional convolutional neural network (CNN) architecture of O’Shea (2016) [1] by selecting the values of the CNN hyperparameters that result in obtaining the best accuracy for each SNR. The optimized model uses CNN4 that increases recognition accuracy of radio modulation over O’Shea’s mode.