Description
A very interesting project that resulted in a paper first submitted to IEEE but currently on revision to increase the soundness of the testing environment.
Our goal was to provide timely feedbacks to teachers thanks to the use of Artificial Intelligence, applying a system that includes GMM modeling and clustering to categorize the classroom speech.
We achieved robust results that converge well to the gold labels set by professional annotators. Our system is able to recognize the four categories: "Teacher Speech", "Student Speech", "Overlapped Speech", and "No speech".
The system has been tested in Chinese classrooms.