DNN-Driven Mixture of PLDA for Robust Speaker Verification

Keyphrases

• Robust Speaker Verification 100%
• Deep Neural Network 100%
• I-vector 100%
• Probabilistic Linear Discriminant Analysis 100%
• Channel Type 60%
• Posterior Probability 40%
• Vector Space 40%
• Mixture Model 40%
• Test Words 40%
• Speaker Verification 20%
• Supervised Learning 20%
• Learning-based 20%
• Discriminative Training 20%
• Indicator Variables 20%
• NIST 2012 SRE 20%
• Marginal Likelihood 20%
• Heterogeneous Cluster 20%
• SNR Mismatch 20%
• Better Performance 20%
• Results-based 20%
• Target Speaker 20%

Computer Science

• Deep Neural Network 100%
• Speaker Verification 100%
• Linear Discriminant Analysis 100%
• Posterior Probability 40%
• Supervised Learning 20%
• Heterogeneous Cluster 20%
• Marginal Likelihood 20%
• Indicator Variable 20%