SilentSpeechClassifier

A set of research tools for silent speech classification, utilising KARA ONE database.

Getting Started

Download the repository, import Dataset and Classifier from SSC.py. I recommend taking a look into the example file. Both Dataset and Classifier are rather well documented.

Prerequisites

• mne (https://martinos.org/mne/stable/index.html)
• Scipy
• Numpy
• Sklearn
• Matplotlib

Known Issues

• grid search goes ape on Windows if started outside "main" and there is parallel computing.
• finding best features works only with Anova and violates the assumption of normality of distribution.

TODO

• plotting

Author

Wojciech Błądek

License

Project is licensed under the MIT License.

Acknowledgments

• Thanks to @stakar (Stanisław Karkosz) for a basis for the feature extraction module.
• Big thanks to the wonderful people behind KARA ONE (http://www.cs.toronto.edu/~complingweb/data/karaOne/karaOne.html)