AI-Powered Silent Speech to Text Transformation

This project introduces a silent speech-to-text system driven by artificial intelligence that uses a web-based interface to translate lip movements into readable text. By combining two crucial preprocessing modules—Frame Blur Removal and Lip ROI Stabilization—the system improves the performance of an already-existing lip-reading model. By identifying and fixing motion-blurred frames, the blur removal module enhances clarity and makes lip contours easier to see. Using facial landmark tracking, the lip ROI stabilization module reduces positional variations brought on by user movement by maintaining a uniform and centered mouth region throughout frames. These improvements give the model cleaner, more reliable input, which increases prediction accuracy. By taking frames from a webcam, processing them on the backend, and displaying the predicted text on the frontend interface, the suggested system runs in real time. According to experimental findings, the additional modules greatly improve the model’s resilience to real-world scenarios like head movement, lighting fluctuations, and camera instability. In addition to supporting future development toward assistive technologies and real-time speechless interaction, the solution shows a workable strategy for silent communication systems.