Speechreading is a difficult task for humans to perform. However, with recent deep learning methods, it is possible to create lip-to-speech systems, which can convert silent lip motion to intelligible audible speech. The task of the student is to learn about recent deep learning methods (e.g. convolutional, recurrent neural networks, transformers, attention) and create new solutions for lip-to-speech conversion. Suggested programming languages: Python. Databases: GRID or UltraSuite-TaL, https://ultrasuite.github.io/data/tal_corpus/

Speechreading is a difficult task for humans to perform. However, with recent deep learning methods, it is possible to create lip-to-speech systems, which can convert silent lip motion to intelligible audible speech. The task of the student is to learn about recent deep learning methods (e.g. convolutional, recurrent neural networks, transformers, attention) and create new solutions for lip-to-speech conversion. Suggested programming languages: Python. Databases: GRID or UltraSuite-TaL, https://ultrasuite.github.io/data/tal_corpus/