Speech Articulation ToolKIT - SATKIT

Tools for processing speech data. While currently these tools mainly work on tongue and larynx ultrasound as well as audio, in future, the toolkit will include facilities for processing other kinds of articulatory data. The first two tools to be implemented are Optical Flow and Pixel Difference.

Optical Flow tracks local changes in ultrasound frames and estimates the flow field based on these.

Pixel Difference and Scanline Based Pixel Difference -- work on raw, uniterpolated data and produce measures of change over the course of a recording. How they work is explained in Chapter 3 of Pertti Palo's PhD thesis. The next stage is a automated reaction time measure, followed possibly by spline distance measures and optic flow.

Major update -- version 1.0 -- coming up

After ICPhS 2023 there will be incremental 0.x releases until the codebase reaches milestones for version 1.0 (See Roadmap). While this notice is here in the main branch, that update is not yet complete. In the meanwhile, if you wish to try SATKIT or contribute to the code or as a beta tester, please contact Pertti for instructions.

With version 1.0 getting closer a lot of the instructions below are outdated. More up-to-date information can be found in the Documentation. See Roadmap for updates on progress towards 1.0.

See Release notes, for what's new.

Getting Started

Download the repository to either a subdirectory of the project you want to use the tools on or a suitable place that you then add to your $PYTHONPATH.

Requisites

SATKit has been tested with the following or more recent:

Python 3.10
PyQt5 v5.12
MatPlotLib 3.5
NumPy 1.22
OpenCV 4.5
SciPy 1.3
praat-textgrids 1.3

A handy way of getting MatPlotLib, NumPy, and SciPy is to get them as part of the Anaconda distribution. Using mamba as the environment manager when creating virtual environments is recommended because conda may not manage to solve the dependencies and may refuse to install anything.

Note: If you decide to use Anaconda, you should install any additional packages you are getting with mamba (or conda) before installing praat-textgrids with pip as the final step of setting up a virtual environment.

For OF you will need in addition:

DIPY 1.5

For processing lip videos:

Scikit-video 1.1

Installing for regular use

An installation through pypi is in the works. Do get in touch if you would like to test it.

A downloadable executable will hopefully also become reality. Do get in touch if you would like to develop it.

Installing for development

To install all of the above and create a conda environment to run SATKIT in first install anaconda from the link above. Then install mamba because anaconda is unlikely to solve the dependencies correctly. Do this on the command line with: conda install mamba -n base -c conda-forge

After you've (possibly forked and) cloned the repository from github with your favourite git tool, cd into the repository root. There you should find the file called satkit_devel_conda_env.yaml and satkit_stable_conda_env.yaml.

Using SATKIT as a library in program development

If you are not going to work on SATKIT itself, you should setup the stable environment. To install the required packages and create the environment, run: mamba env create -f satkit_stable_conda_env.yaml

Activate the environment with conda activate satkit-stable and the basics are set up. If you want to have an editable install (changing code in the repo changes behaviour of the installed SATKIT immediately) run finally pip install -e . or python3 -m pip install --editable . if you prefer being explicit.

Working on SATKIT

If you plan on working on SATKIT, run the commands mamba env create -f satkit_devel_conda_env.yaml and conda activate satkit-devel instead of the ones above.

What's included

See MANIFEST.markdown for a description of the contents.

Running the examples

There are three small datasets included in the distribution. You can run tests on them with the test script pd_test.py. Currently the following work and produce a new spaghetti_plot.pdf and a transcript in [method_name].log.

python satkit.py tongue_data_1_1

The first example directory contains recordings with all files present while the second is intentionally missing some files. The latter case should therefore produce warnings in the resulting log. Running without the exclusion list specified should produce a plot with a couple more curves in it.

The routines to deal with a directory structure like that of test2 are yet to be implemented.

Running the tests

Proper testing is yet to be implemented.

Contributing

Please, read the docs and get in touch with Pertti, if you would like to contribute to the project.

Versioning

We use SemVer for versioning under the rules as set out by PEP 440 with the additional understanding that releases before 1.0 (i.e. current releases at time of writing) have not been tested in any way.

For the versions available, see the tags on this repository.

Authors

Pertti Palo - Initial work on PD and the Python framework used by the combined toolkit - giuthas
Scott Moisik - Initial work on OF - ScottMoisik
Matthew Faytak - mfaytak
Motoki Saito - Producing interpolated ultrasound images from raw data msaito8623

List of contributors will be updated once there are more people working on this project.

Copyright and License

The Speech Articulation ToolKIT (SATKIT for short) and examples is a tool box for analysing articulatory data.

Program license

SATKIT is licensed under GPL 3.0.

This program (see below for data) is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see https://www.gnu.org/licenses/gpl-3.0.en.html

Data license

Data License

The data in directories larynx_data, tongue_data_1, tongue_data_1_2, and tongue_data_2 are licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0) License. See link above or https://creativecommons.org/licenses/by-nc-sa/4.0/ for details.

Citing the code

When using any part of SATKIT, please cite:

Palo, P., Moisik, S. R., and Faytak, M. (2023). “Analysing Speech Data with SATKIT”. In: International Conference of Phonetic Sciences (ICPhS 2023). Prague.
Faytak, M., Moisik, S. & Palo, P. (2020): The Speech Articulation Toolkit (SATKit): Ultrasound image analysis in Python. In ISSP 2020, Online (planned as Providence, Rhode Island)

When making use of the Optic Flow code, please cite:

Esling, J. H., & Moisik, S. R. (2012). Laryngeal aperture in relation to larynx height change: An analysis using simultaneous laryngoscopy and laryngeal ultrasound. In D. Gibbon, D. Hirst, & N. Campbell (Eds.), Rhythm, melody and harmony in speech: Studies in honour of Wiktor Jassem: Vol. 14/15 (pp. 117–127). Polskie Towarzystwo Fonetyczne.
Moisik, S. R., Lin, H., & Esling, J. H. (2014). A study of laryngeal gestures in Mandarin citation tones using simultaneous laryngoscopy and laryngeal ultrasound (SLLUS). Journal of the International Phonetic Association, 44(01), 21–58. https://doi.org/10.1017/S0025100313000327
Poh, D. P. Z., & Moisik, S. R. (2019). An acoustic and articulatory investigation of citation tones in Singaporean Mandarin using laryngeal ultrasound. In S. Calhoun, P. Escudero, M. Tabain, & P. Warren (Eds.), Proceedings of the 19th International Congress of the Phonetic Sciences.

When using the Pixel Difference (PD) code, please cite:

Pertti Palo (2019). Measuring Pre-Speech Articulation. PhD thesis. Queen Margaret University, Scotland, UK. Available here PhD thesis.

Acknowledgments

Inspiration for PD was drawn from previous projects using Euclidean distance to measure change in articulatory speech data. For references, see Pertti Palo's PhD thesis.
The project uses a nifty python tool called licenseheaders by Johann Petrak and contributors to add and update license headers for python files.

giuthas / satkit