edward-reehorst / RadioML_2016

Open RadioML Synthetic Benchmark Dataset

RadioML 2016

Fork of adeeb10abbas/RadioML_2016, which is a fork of radioML/dataset

Goal

Remove dependencies on gr-mapper and gr-mediatools

Dependencies

• python
• gnuradio
• numpy
• scipy

Setup

conda create -n gnuradio

conda activate gnuradio

conda config --env --add channels conda-forge

conda config --env --set channel_priority strict

conda install numpy scipy gnuradio=3.8

Create Dataset

First, create an output folder

mkdir datasets

There are two constants in generate_RML2016.10a.py:

• apply_channel - whether or not to model the dynamic transmission channel
• thermal_noise - whether or not to model AWGN at the receiver Set these parameters to your preference.

Then run:

python generate_RML2016.10a.py

Changes

• We are using wavfile source instead of gr-mediatools.
• We are using built in modulations instead of gr-mapper
• Chnage analog source from serial podcast (which has a lot of silence) to public domain recording of The Star Spangled Banner (which was editted to remove silence at the beginning of the recording).
• Add option to turn off AWGN