This project aims to find the optimal model for classifying modulation schemes using deep learning techniques. The implemented models include:
- CNN
- ResNet
- LSTM
- CNN-LSTM
- Inception Network
- CNN-LSTM with attention networks
- VGG
- Dual Stream LSTM-CNN Network
- Dual Stream CNN.
The following dependencies are required to run this project:
- Keras
- Tensorflow
- Python
- Matplotlib
- Numpy
- Pandas
The dataset used in this project is provided by DeepSig.ai and it is called RML2016.10a. The dataset contains 11 different modulation schemes and the data is generated using a software-defined radio (SDR) platform. It is widely used in the wireless communications research community as a benchmark dataset for evaluating the performance of different modulation classification techniques.
To run the project, you'll need to download the RML2016.10a dataset from DeepSig.ai and place it in your local directory. Then, clone the repository and run the main file in the root directory. The results of the different models and the optimal model will be displayed.
Note: Make sure to have all the dependencies installed before running the project.
This project presents a study of the performance of various deep learning models on the RML2016.10a dataset for modulation classification, and the results demonstrate that the optimal model for this task is Dual Stream CNN.
We hope this project will be useful for researchers and practitioners working in the field of wireless communications and modulation classification.
Please feel free to reach out to me if you have any questions or issues with the project.