基於視覺的手語辨識與語音轉換系統以促進包容性溝通

// A vision-based sign language recognition & voice conversion system for inclusive communication

本研究提出結合卷積神經網絡與Transformer分類器的手語轉語音系統｜建立多樣化手勢資料集並採用先進預處理提升辨識效果｜系統準確率達99%，優於傳統方法，具備即時且穩定的輔助溝通價值，特別適用於聽障及語言障礙者緊急情況下使用。

// This study presents a vision-based sign-to-voice system using convolutional neural networks combined with transformer-based classification to recognize hand gestures. A custom dataset with varied conditions was created, and advanced preprocessing steps were applied to improve gesture representation. The system achieved high accuracy (99%) and reliability, outperforming traditional methods. This robust approach offers a practical, real-time assistive communication tool for deaf and speech-impaired individuals, especially in emergencies.

Published 2026年7月8日