Deep Binaural Spatio-Temporal Wiener Filtering

This repository contains supplementary code for the paper "Imposing Correlation Structures for Deep Binaural Spatio-Temporal Wiener Filtering" by M. Tammen and S. Doclo, IEEE Trans. Audio, Speech and Language Processing, vol. 33, pp. 1278-1292, 2025.

Installation

1. Clone this repository to your machine.
2. Make sure Anaconda or Miniconda are available.
3. Create and activate the conda environment using the provided environment.yml file. We recommend using Mamba for a faster installation.

   # If you don't have mamba, install it first:
   # conda install -n base -c conda-forge mamba
   mamba env create -f environment.yml
   conda activate j2

Usage

You can use the inference.py script to enhance a noisy audio file using one of the pretrained models mentioned in the paper.

Example

To run the inference script, use a command like the following:

python inference.py --model stwf_noCommonSTCM_noRTF --input data/noisy.wav --output data/noisy_enhanced.wav

This command will:

• Load the stwf_noCommonSTCM_noRTF model.
• Process the data/noisy.wav file.
• Save the enhanced audio to data/noisy_enhanced.wav.

If you don't specify an output file, the enhanced audio will be saved in the same directory as the input file with _enhanced appended to the name.

Available Models

The following models are available for use with the --model argument (see Table II in the paper):

• stwf_noCommonSTCM_noRTF
• stwf_CommonSTCM_noRTF
• stwf_CommonSTCM_globalRTF
• stwf_CommonSTCM_ipsiRTF
• stwf_bilat_CommonSTCM_noRTF
• stwf_bilat_CommonSTCM_global
• df_noRTF