paderwasn

Paderwasn is a collection of methods for acoustic signal processing in wireless acoustic sensor networks (WASNs).

Installation

Install requirements:

$ pip install --user git+https://github.com/fgnt/lazy_dataset.git@ce8a833221580242e69d43e62361adca02478f79
$ pip install --user git+https://github.com/fgnt/paderbox.git@7fed5b44be2effcedb7a26778ada6c5668b1d6bd

Clone the repository:

$ git clone https://github.com/fgnt/paderwasn.git

Install package:

$ pip install --user -e paderwasn

Content

Algorithms:
- Geometry calibration:
  - Geometry calibration using iterative data set matching [1]
  - GARDE-algorithm [2]
- Signal synchronization:
  - Sampling rate offset (SRO) estimation:
    - Dynamic weighted average coherence drift (WACD) [3]
    - Onlne WACD [4]
  - Sampling time offset (STO) estimation [3]
  - Resampling to compensate for an SRO
  - Simulation of a (time-varying) SRO [3]
Databases:
- Geometry calibration observations: Collection of direction-of-arrival (DoA) and source-node distance estimates used for geometry calibration in [1]
- Asynchronous WASN database: Database of simulated audio signals which were recorded by an asynchronous WASN. This database corresponds to the database (after minimal adjustments) used in [3] for evaluation of signal synchronization algorithms.
Experiments using the provided algorithms and databases:
- Comparision of geometry calibration methods
- Comparision of SRO methods
- STO estimation

Asynchronous WASN database

The ansynchronous WASN database consists of simulated recordings of a asynchronous WASN with four sensor nodes. The database corresponds to a slightly modified version of the database used in [3] (Source signals stemming from the Timit datbase were replaced by signals stemming from the LibriSpeech database). Four scenarios are simulated (see [3] for details):

Scenario	Time-varying SRO	Multiple Source Positions	Speech Pauses
Scenario-1
Scenario-2	X
Scenario-3	X	X	X
Scenario-4	X	X

To prepare the databse follow these steps:

Download the room impulse responses (RIRs), generated by the generator of Habets. using this python port, SRO trajectories (see [3]) and simulation descriptions:
```
$ python -m paderwasn.databases.synchronization.download with 'database_path="/PATH/WHERE/TO/STORE/THE/DATABASE/"'
```
If you do not have downloaded the LibriSpeech database (test-clean) before download the test-clean part of LibriSpeech:
```
$ python -m paderwasn.databases.synchronization.download with 'database="librispeech"' 'database_path="/PATH/WHERE/TO/STORE/THE/DATABASE/"'
```

Create a json-file for the database:

$ python -m paderwasn.databases.synchronization.create_json with 'database_path="/PATH/TO/THE/DATABASE/"' 'librispeech_path="/PATH/TO/THE/ROOT/OF/LIBRISPEECH/"' 'json_path="/PATH/WHERE/TO/STORE/THE/DB_JSON/"'

Create a file-based version of the database, i.e. simulate the audio signals, store the audio signals on the disk and create a new json-file for the file-based version of the database:

$ python -m paderwasn.databases.synchronization.write_files with 'json_path="/PATH/TO/THE/DB_JSON/"' 'data_root="/PATH/WHERE/TO/STORE/THE/FILE_DB/"' 'json_file_db_path="/PATH/WHERE/TO/STORE/THE/FILE_DB_JSON/"'

References

[1] Gburrek, T., Schmalenstroeer, J., Haeb-Umbach, R.: "Geometry Calibration in Wireless Acoustic Sensor Networks Utilizing DoA and Distance Information", EURASIP Journal on Audio, Speech, and Music Processing, vol. 2021, no. 1, pp. 1–17, 2021.

[2] Gburrek, T., Schmalenstroeer, J., Haeb-Umbach, R.: "Iterative Geometry Calibration from Distance Estimates for Wireless Acoustic Sensor Networks". in Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2021, pp. 741-745.

[3] Gburrek, T., Schmalenstroeer, J., Haeb-Umbach, R.: "On Synchronization of Wireless Acoustic Sensor Networks in the Presence of Time-varying Sampling Rate Offsets and Speaker Changes". Submitted to IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2022, arXiv preprint arXiv:2110.12820.

[4] Chinaev, A., Enzner, G., Gburrek, T., Schmalenstroeer, J.: “Online Estimation of Sampling Rate Offsets in Wireless Acoustic Sensor Networks with Packet Loss,” in Proc. 29th European Signal Processing Conference (EUSIPCO), 2021, pp. 1–5.

Citation

If you are using the code or one of the provided databases please cite the corresponding paper (If you use the asynchronous WASN database please cite [3]):

@article{gburrek2021geometry,
   title={Geometry calibration in wireless acoustic sensor networks utilizing DoA and distance information},
   author={Gburrek, Tobias and Schmalenstroeer, Joerg and Haeb-Umbach, Reinhold},
   journal={EURASIP Journal on Audio, Speech, and Music Processing},
   volume={2021},
   number={1},
   pages={1--17},
   year={2021},
   publisher={Springer}
}

@inproceedings{gburrek2021synchronization, 
   author={Gburrek, Tobias and Schmalenstroeer, Joerg and Haeb-Umbach, Reinhold}, 
   booktitle={IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
   title={Iterative Geometry Calibration from Distance Estimates for Wireless Acoustic Sensor Networks},
   year={2021},
   pages={741-745},
   doi={10.1109/ICASSP39728.2021.9413831}
}

@misc{gburrek2021synchronization,
     title={On Synchronization of Wireless Acoustic Sensor Networks in the Presence of Time-varying Sampling Rate Offsets and Speaker Changes}, 
     author={Gburrek, Tobias and Schmalenstroeer, Joerg and Haeb-Umbach, Reinhold},
     year={2021},
     eprint={2110.12820},
     archivePrefix={arXiv},
     primaryClass={eess.AS}
}

@inproceedings{Chinaev2021,
   author = {Chinaev, Aleksej and Enzner, Gerald and Gburrek, Tobias and Schmalenstroeer, Joerg},
   booktitle = {29th European Signal Processing Conference (EUSIPCO)},
   pages = {1--5},
   title = {{Online Estimation of Sampling Rate Offsets in Wireless Acoustic Sensor Networks with Packet Loss}},
   year = {2021},
}

Acknowledgment

Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - Project 282835863 (Deutsche Forschungsgemeinschaft - DFG-FOR 2457).

YoshikiMas / paderwasn