From: Deep neural network-based bottleneck feature and denoising autoencoder-based dereverberation for distant-talking speaker identification
Values
Number of layers
9
Number of units in each layer
1,024
Context size
9 (p4 + c1 + l4)
