Modulation Classification

Table of contents

• Description
• Model structure
• Data in this project
• Files in this repo
• Results
• Dependencies for this project
• Running the code

Description

This project implements two deep learning (DL) models on classifying the modulation formats of various radio frequency (RF) signals. The two DL models are the Artificial Neural Network (ANN) and Convolutional Neural Network (CNN), while the radio frequency signals are generated by GNU Radio software and consist of 11 modulation formats (8 digital and 3 analog) at varying signal-to-noise ratios (SNR). Conducting modulation classification on the RF signals, this work compares the performance of ANN and CNN models at different values of SNR. Plus, this project applies Principal Component Analysis (PCA) technique on both ANN and CNN models. The purpose is to observes the trade-off relationship between the classification accuracy and the length of training time

Model structure

ANN	CNN

Data in this project

The original dataset in this project is publicly available on the DeepSig RF Datasets For Machine Learning. The dataset is called RADIOML 2016.10A and is generously offered by DeepSig Inc.

Files in this repo

This GitHub repo contains the following folders or files

• img folder contains the images of the training results, neural network structures, and test results

• model folder contains all the trained ANN and CNN models used in this project

• ANN_vs_CNN.ipynb is the code for comparing the performance of ANN and CNN on classifying the modulation schemes of the RF signals in dataset

• ANN_using_PCA.ipynb and CNN_using_PCA.ipynb are the code for applying PCA on the original dataset and observing the performance of ANN and CNN versus different values of dimension reduction

• ANN.ipynb and CNN.ipynb are the prototype of ANN and CNN models at the beginning stage of this project

• Project_Report.pdf is the report of this project, which contains specific technical details and in-depth discussion

Results

Comparison between ANN and CNN

 

Comparison between ANN and CNN with PCA

ANN	CNN

Dependencies for this project

This project requires the following python modules:

numpy  matplotlib.pyplot  pickle  sklearn  random  tensorflow.keras  seaborn  time

Please make sure you have all the modules installed before running the code. For installing these modules, one can use command pip install or conda install

Running the code

1. Click here to download the dataset for this project. It may require contact verification before downloading the dataset

2. Extract the dataset from the zipped folder into a folder named data

3. Download this repository to your local machine

git clone https://github.com/zhaoshengEE/Modulation_Classification.git

4. Relocate the data folder into the repository

5. Start your journey on this project