This repo contains modules and scripts to train a modulation classifier and an implementation for a software defined radio using GNU Radio for real-time modulation classification. The module extract_features.py contains functions to load IQ data, augment it with AWGN, extract features from it, and convert it into a format for training a classifier. The module features.py contains various feature extraction methods implemented from publications in automatic modulation classification. The file train_model.py trains either a decision tree or a feed forward neural network with the specified features and augments the training data with AWGN with specified SNR levels. The performance of the classifier is displayed using the functions in the plots.py module. The classifier may then be tested with different SNR levels with the script test_model.py. This uses an implementation of the forward pass of a feedforward neural network in FeedForwardNetwork.py. This implementation of a neural network is used in the GNU Radio script modulation_classifier.py which is a real-time simulation of the modulation classifier for a software defined radio.

• NumPy
• SciPy
• Matplotlib
• Scikit-Learn
• GNURadio