End-To-End Audiovisual Feature Fusion for Active Speaker Detection

Keyphrases

• Feature Extracting 100%
• Time Application 100%
• Temporal Dynamics 100%
• Feature Extraction Methods 100%
• Feature Fusion 100%
• ConvNet 100%
• Bidirectional Gated Recurrent Unit (BiGRU) 100%
• Active Speaker Detection 100%
• Inference Time 100%
• Audiovisual Features 100%
• Audio 50%
• Experiment Results 50%
• Visual Input 50%
• Detection Result 50%
• Two-stream 50%
• Human-machine Interaction 50%
• Model Inference 50%
• Noise Signal 50%
• Mel Frequency Cepstral Coefficients 50%
• Audio Input 50%
• Similar Features 50%
• Input Size 50%

Computer Science

• Feature Extraction 100%
• Extracted Feature 100%
• Real-Time Application 100%
• Temporal Dynamic 100%
• Feature Fusion 100%
• Human-Computer Interaction 50%
• Detection Result 50%
• Inference Model 50%