Code for the paper: Changhong Wang, Emmanouil Benetos, Shuge Wang, and Elisabetta Versace. Joint Scattering for Automatic Chick Call Recognition, IEEE European Signal Processing Conference (EUSIPCO), 2022.

git clone https://github.com/changhongw/jointscat-chickcall.git

conda create -n myenv python=3.7.13
conda install --file requirements.txt

git clone https://github.com/changhongw/scatnet.git

The code is based on ScatNet, a MATLAB implementation of the scattering transform, with monior modifications. We can run scatnet in Python using subprocess module (see JTFS feature extraction below).

You can run directly on your own audio data or use the chick call data in the paper. The chick call dataset is available for downloading. The recordings being used are listed in file_names.txt.

python JTFS_feature_extract.py

• chick call detection:
  Detection_onset_segment.ipynb

• chick call recognition:
  Scat_Only.ipynb
Seg_Scat_annotated.ipynb
Seg_Scat_detected.ipynb
MFCC_Only.ipynb
Seg_MFCC_annotated.ipynb
Seg_MFCC_annotated.ipynb
  

@inproceedings{wang2022chickcall,
  title={Joint Scattering for Automatic Chick Call Recognition},
  author={Wang, Changhong and Benetos, Emmanouil and Wang, Shuge and Versace, Elisabetta},
  booktitle={IEEE European Signal Processing Conference (EUSIPCO)},
  year={2022}
}