SignBridge — Two-Way Filipino Sign Language Communication System
> Role: Lead Researcher & Machine Learning EngineerBuilt an end-to-end Windows desktop workstation in Python and PyQt5 bridging communication barriers between Deaf or hard-of-hearing citizens and Local Government Unit (LGU) frontline personnel through computer vision recognition, multilingual speech synthesis, and video generation.
Features a real-time FSL recognition pipeline extracting 162 hand and upper-body landmark coordinates per frame with MediaPipe, processed through an LSTM sequence classifier (30-frame window). Dynamically assembles sign videos with FFmpeg and provides local speech-to-text transcription for Filipino, English, and Taglish.
- Model Architecture: LSTM Neural Network
- Input Vector: 30 frames × 162 features
- Landmark Sensor: MediaPipe Hands + Pose
- Video Assembler: Dynamic FFmpeg Pipeline
- Supported Dialects: Tagalog · English · Taglish
- Deployment: Desktop GUI (PyQt5)