AI & computer vision
Egyptian Arabic Lip-reading
My graduation project at Nile University, exploring visual speech recognition for Egyptian Arabic as part of a research team.
Project overview
For my 2024 graduation project at Nile University, I worked with a team to explore lip-reading for Egyptian Arabic. The broader research considered how a sequence of visual speech frames could support recognition of spoken language.
Engineering problem
Mouth motion is a small part of a video frame. A useful visual input pipeline needs to locate that region and follow it through a sequence while limiting irrelevant image content. Tracking quality matters because downstream recognition depends on the frames it receives.
My contribution
I developed the mouth-tracking component within the team project. My work focused on identifying the mouth region and preparing visual input for the wider research pipeline. The full recognition system was a team effort.
Technical approach
The existing project record identifies Python, MediaPipe and OpenCV for the computer vision work. Face landmarks provide a basis for locating the mouth region, which can then be isolated from successive video frames.
The illustration on this page explains that preprocessing concept. It is not an output from the research model, an internal screenshot or a measured result.
Results & research status
The work formed part of my 2024 graduation project. This case study documents my component-level contribution. A trained model, reproducible evaluation and public research repository are not included in this portfolio.