PL-EESR: Perceptual Loss Based End-to-End Robust Speaker Representation Extraction

Keyphrases

• Background Noise 50%
• Clean Environment 100%
• Deep Learning Framework 50%
• Embedding Extraction 50%
• End-to-end Deep Learning 50%
• Identification Task 50%
• Loss-based 100%
• Noisy Environment 100%
• Perceptual Deviation 50%
• Perceptual Loss 100%
• Perceptual Quality 50%
• Relative Information 50%
• Speaker Embedding 50%
• Speaker Identification 50%
• Speaker Information 50%
• Speaker Representation 100%
• Speaker Verification 50%
• Speech Distortion 50%
• Speech Enhancement 50%
• Speech Signal 100%
• Verification Task 50%

Computer Science

• Background Noise 100%
• Deep Learning Method 100%
• Information Loss 100%
• Learning Framework 100%
• Noisy Environment 100%
• Perceptual Quality 100%
• Speaker Identification 100%
• Speaker Verification 100%
• Speech Enhancement 100%