Efficient Audiovisual Fusion for Active Speaker Detection

Keyphrases

• Active Speaker Detection 100%
• Audio-visual Fusion 100%
• Fusion Approach 60%
• Computation Overhead 40%
• Feature Dimension 40%
• Detection Accuracy 40%
• Performance Enhancement 20%
• Audio 20%
• Multiple Stages 20%
• Facial Features 20%
• Fusion Process 20%
• Sound Signal 20%
• Detection Performance 20%
• Fusion Technique 20%
• Smoothing Methods 20%
• Unsynchronized 20%
• Facial Regions 20%
• Feature Fusion 20%
• Validation Set 20%
• Fused Features 20%
• Ablation Experiment 20%
• Complex Technique 20%
• Visible Human 20%
• Speaker Change Detection 20%
• Competitive Detection 20%
• Audio Features 20%
• Image Scale 20%
• Video Module 20%
• Multiplex Networks 20%
• Noise Signal 20%
• Intramodal 20%

Computer Science

• Postprocessing 100%
• Feature Dimension 66%
• Detection Accuracy 66%
• Enhance Performance 33%
• Fusion Technique 33%
• Detection Performance 33%
• Feature Fusion 33%
• Validation Set 33%
• Facial Feature 33%
• Concatenation Operation 33%
• Multiple Network 33%