Automatic Emoticons Insertion System Based on Acoustic Information of User Voice: 4th Report on Silence Removal for Generalized Emotion Estimation and Improved Visual Representations

This study proposes an automatic emoticon insertion system based on acoustic information from user voice. The system estimates emotions from acoustic features using Support Vector Machine (SVM) and inserts corresponding emoticons to enhance emotional communication in text-based social media. Key advancements include the removal of silent segments from training model voice data, improving emotion estimation accuracy and generalizability, as well as a refined visual representation method to address the polysemous nature of emoticons. Experimental results show a significant performance improvement, with accuracy rising to 75.12% compared to the previous model. These findings lay the groundwork for real-time applications, supporting enhanced user communication efficiency and emotion expressiveness in digital contexts.