Copyright 2022 Zhengjun Yue, Heidi Christensen, Jon Barker

This is a Pytorch-Kaldi recipe to build automatic speech recognition systems on the Torgo corpus of dysarthric speech. "Multi-modal Acoustic-articulatory Feature Fusion for Dysarthric Speech Recognition"

Run the following:

. ./env_setup.sh

. ./cmd.sh
. ./path.sh
. ./parse_options.sh

The following instructions allow to train ASR systems on Torgo and to reproduce results from the paper.

# MFCC baseline
python -u run_exp.py cfg/fd1_mfcc.cfg 

# Multi-stream fusion
python -u run_exp.py cfg/fd1_mfcc_concat2.cfg 

Please cite the following paper if you use this script for your research or are interested in this paper.

@inproceedings{yue2022multi,
  title={MULTI-MODAL ACOUSTIC-ARTICULATORY FEATURE FUSION FOR DYSARTHRIC SPEECH RECOGNITION},
  author={Yue, Zhengjun and Loweimi, Erfan and Cvetkovic, Zoran and Christensen, Heidi and Barker, Jon},
  booktitle={ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing},
  year={2022}
}

The code is based on an earlier recipe by Mirco Ravanelli, Titouan Parcollet and Yoshua Bengio.