Skip to main content

Advertisement

Table 7 Comparison of results of artificial data with different compensation parameter sets and combination methods for speaker identification

From: Distant-talking speaker identification by generalized spectral subtraction-based dereverberation and its efficient computation

  Number of impulse response  
  condition for test (%) Average (%)
  4 5 6 7 8
Parameters (n, α)       
(0.1, 0.1) 95.2 90.2 87.2 90.5 92.6 91.1
(0.3, 0.3) 96.2 89.9 89.6 87.8 89.9 90.7
(0.5, 0.5) 96.2 91.1 91.4 90.1 92.3 92.2
(0.7, 0.7) 95.0 90.3 91.4 90.8 92.5 92.0
(1.0, 1.0) 94.0 90.6 91.0 90.5 92.3 91.7
(0.1, 0.2) 94.6 88.9 88.0 86.0 90.4 89.6
(0.3, 0.6) 95.8 89.9 89.8 88.8 91.3 91.1
(0.5, 1.0) 95.3 91.0 91.0 90.5 93.1 92.2
(0.7, 1.4) 94.5 90.9 91.7 90.6 92.9 92.1
(1.0, 2.0) 93.8 89.8 90.9 90.3 92.0 91.3
Conventional       
combination 96.2 92.5 92.6 92.5 94.1 93.6
Efficient       
combination 96.2 92.5 92.6 92.5 94.1 93.6
(N=5)