ICASSP 2024 SP Grand Challenge:

Data-Driven Signal Separation in Radio Spectrum

This challenge will require developing an engine for signal separation of radio-frequency (RF) waveforms. At inference time, a superposition of a signal of interest (SOI) and an interfering signal will be fed to the engine, which should recover the SOI by performing a sophisticated interference cancellation. SOI is a digital communication signal whose complete description is available (modulation, pulse-shape, timing, frequency, etc). However, the structure of the interference will need to be learned from data. We expect successful contributions to adapt existing machine learning (ML) methods and/or propose new ones from the areas of generative modeling, variational auto-encoders, U-Nets and others.

Final leaderboard (the last column of each table shows the results averaged over the 8 mixture cases studied in this challenge):

Important Dates:

  • Oct. 4, 2023 – Submission 1 deadline: Initial submission containing outcomes on TESTSET1MIXTURE.
  • Nov. 1, 2023 – Submission 2 deadline: Second submission containing the outcomes on TESTSET1MIXTURE.
  • Dec. 1, 2023 Dec. 8, 2023 – Final submission deadline: Last submission containing the outcomes on TESTSET2MIXTURE. The final ranking will be exclusively determined by the results of this ultimate submission.
  • Jan. 2, 2024 Jan. 9, 2024 – Submission deadline for 2-page papers of the best Challenge submissions (by invitation only). Accepted papers will be in the ICASSP proceedings.

Further information and details can be found below.


Challenge Details

Click here for details on the challenge setup


Click here to access to access the challenge discord server


Link to dataset

Link to TestSet1Mixture

Starter Code

Click here for the starter code of this challenge

Reference Methods

Click here for the Jupyter notebook of selected reference methods

Link to baseline model weights

Papers about this challenge (to appear in the ICASSP24 proceedings):

Challenge organizers’ paper: The Data-Driven Radio Frequency Signal Separation Challenge

KU-TII paper: A Novel Approach to Wavenet Architecture for RF Signal Separation with Learnable Dilation and Data Augmentation

OneInAMillion paper: Signal separation in radio spectrum using self-attention mechanism

LHen paper: Improving Data-Driven RF Signal Separation with Soi-Matched Autoencoders

TUB paper: Demucs for Data-Driven RF Signal Denoising

imec_DLab paper: A U-Net Architecture for Time-Frequency Interference Signal Separation of RF Waveforms

DowntotheWires paper: Optimized Size-Performance Model for Interference Rejection in Digital Communications for the ICASSP 2024 Challenge

* The intellectual property (IP) is not transferred to the challenge organizers; in other words, if code is shared or submitted, the participants retain ownership of their code.