We implement CNN-based Automatic Modulation Classification (AMC) by applying a polar transformation to the received symbols along with a modification to the kernel dimensions of the CNN in order to improve the performance of the classification procedure under phase imperfections. This repository is part of published research paper in IEEE Wireless Communications Letters with title "CNN-Based Automatic Modulation Classification under Phase Imperfections".
The radioml.m file is a Matlab file that can generate various of constellation datasets affected by phase noise and AWGN. You can use it us guidline to create your own datasets.
The file ex.py implements both LeNet and AlexNet and evaluates the performance of conventional methods and the proposed method.
To download the dataset that we utilized in our published paper you can navigate to url https://drive.google.com/file/d/1_uQczmLe_B2damm1fCBaCC2n6w0YVyg4/view?usp=drive_link in order to download the datasets.zip file.