From: Deep neural network-based bottleneck feature and denoising autoencoder-based dereverberation for distant-talking speaker identification
Method
Number of layers
CENSREC-4 database
Ave.
0.40
0.60
0.75
DAE
1
95.35
90.65
95.20
93.73
3
95.40
91.50
94.70
93.87
5
94.30
90.85
93.80
92.98
7
89.65
84.40
87.60
87.22
9
87.85
82.75
87.50
86.03