On Reduction and Optimization of AI Parameter Set via Signal Processing Methods: Lip-Sync Problem
33rd Conference on Signal Processing and Communications Applications-SIU-Annual, İstanbul, Turkey, 25 - 28 June 2025, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/siu66497.2025.11111772
- City: İstanbul
- Country: Turkey
- Istanbul University Affiliated: No
Abstract
Lip synchronization is a fundamental component of speech-driven applications, ranging from virtual reality and human-machine interaction to forensic analysis and cybersecurity. Traditional signal processing-based methods for lip synchronization face challenges in accuracy, real-time performance, and language independence. This study benchmarks a statistical signal processing-based lip synchronization approach against modern machine learning-based tools. By leveraging high signal-to-noise ratio (SNR) audio data and transcripts, we enhance the conventional algorithm with AI-driven models. The integration of signal processing with machine learning and deep learning enables more precise, natural, and language-agnostic lip synchronization. The results emphasize AI's transformative impact on speech processing technologies and set the stage for future advancements in multimodal communication.