Development and Evaluation of SALIN: A Context-Aware Real-Time Filipino Sign Language Translation System Using Gesture and Facial Cues

This study presents SALIN, a context-aware real-time Filipino Sign Language (FSL) translation system that integrates hand gesture recognition with facial cue analysis to improve phrase-level interpretation. The system employs a multimodal vision-based framework with temporal stability controls to prevent looping and fluctuating outputs during continuous translation. SALIN was evaluated under live webcam translation and offline video analysis scenarios using both gesture-only and multimodal configurations. Experimental results show that the multimodal approach achieved higher translation accuracy (88.9%) compared to the gesture-only baseline (78.4%), while maintaining real-time performance with an average processing latency of approximately 150 ms. Confidence-based commitment and temporal smoothing ensured that most translations were produced at high confidence levels, supporting stable and reliable output. The findings demonstrate that incorporating facial cues and context-aware modeling significantly enhances translation accuracy without compromising responsiveness, indicating the suitability of SALIN for practical assistive communication and accessibility applications.