Skip to main content

Table 16 Three different training sets used to train the DNN and GMM

From: Deep neural network-based bottleneck feature and denoising autoencoder-based dereverberation for distant-talking speaker identification

Training DAE and BF-DNN GMM training
set training data data
number   
1 5 (utterances) × 100 (speakers) 5 (utterances) ×100 (speakers)
  × 3 (environments) × 3 (environments)
2 Same as training 10 (utterances) ×100 (speakers)
  set 1 × 3 (environments)
3 10 (utterances) ×100 (speakers) Same as training
  × 3 (environments) set 2