SATCOM Interference Classification Dataset

SATCOM Interference Dataset and Generation Script This repository contains the dataset and MATLAB generation script for the paper: "Lightweight Dual-Branch CRNN for Interference Classification in SATCOM". The goal of this work is to provide a realistic, high-fidelity synthetic dataset for developing and benchmarking machine learning models for interference classification in satellite communications. Dataset Overview The provided dataset "satcom_interference_dataset_92400inst.mat" contains 92,400 signal instances, each representing a 20 µs snapshot captured at a sampling rate of 50 Msps. The dataset is built around a key feature for realism: imperfect signal cancellation. Instead of assuming the desired DVB-S2X signal is perfectly removed, a low-power, distorted residual is always present. All signals are also passed through a receiver model that introduces Carrier Frequency Offset (CFO) and Phase Noise. Interference Classes The dataset includes 7 classes (with labels 0-6): Label Class Name Description 0 NoInterference Contains only the DVB-S2X residual, thermal noise, and receiver impairments. 1 CW A single-tone Continuous Wave jammer. 2 FMCW A linear frequency-modulated (chirp) jammer. 3 Pulsed A simple pulsed jammer with varying duty cycle and pulse width. 4 BroadbandNoiseJammer A non-stationary, band-limited noise jammer with amplitude modulation. 5 5G_NR_OFDM A realistic 5G NR OFDM signal from a terrestrial base station. 6 AdjacentDVB_S2X Interference from a neighboring satellite's DVB-S2X signal. How to Use the Pre-Generated Dataset The primary dataset file used in the paper is:"satcom_interference_dataset_92400inst.mat" You can load it into MATLAB using the following command: % Load the dataset fileload('satcom_interference_dataset_92400inst.mat'); % The file contains the following variables:% - all_iq_data: [92400 x 1000] complex single matrix of IQ samples.% - all_labels: [92400 x 1] uint8 vector of class labels (0-6).% - all_metadata: [92400 x 1] struct array with detailed parameters for each instance.% - fs, N_points, etc.: Scalar parameters describing the dataset generation. Generating a Custom Dataset If you wish to generate a new dataset with different parameters (e.g., more instances, different CNR/INR ranges), you can use the provided generation script. Requirements MATLAB R2022a or newer Required Toolboxes: Communications Toolbox Signal Processing Toolbox Phased Array System Toolbox 5G Toolbox Parallel Computing Toolbox Setup DVB-S2X LDPC Matrices: The script requires LDPC parity matrices for DVB-S2X waveform generation. Download the support package from the MathWorks DVB-S2X Library page. After installation, locate the s2xLDPCParityMatrices folder. Copy this folder into the same directory as the generate_satcom_dataset.m script. The script is configured to look for it in a relative path (./s2xLDPCParityMatrices). Configuration and Execution Open generate_satcom_dataset.m in MATLAB. Edit the "User Configuration" section at the top of the script to set your desired parameters (e.g., numInstancesPerClass, CNR_dB_range, INR_dB_range). Run the script from the MATLAB command window: run('generate_satcom_dataset.m'); The script will create a new folder named satcom_interference_dataset and save the resulting .mat file inside it. Generation can take a significant amount of time depending on the number of instances and your hardware. Citation If you use this dataset or generation script in your research, please cite our paper: @inproceedings{agamy2025Lightweight, author = {Agamy, Abdallah and Naguib, Mohamed and Abdelwanees, Essam and Rohiem, Alaa Eldin}, title = {Lightweight Dual-Branch CRNN for Interference Classification in SATCOM}, booktitle = {2025 37TH THE INTERNATIONAL CONFERENCE ON MICROELECTRONICS (ICM)}, year = {2025}, pages = {1-6}, doi = {10.1109/ICM66518.2025.11322483}}