Skip to main content

Advertisement

Table 2 Female speaker verification results of GMM-UBM method in terms of percent EER (minDCF) for the proposed algorithm, Drugman’s VAD method [27], and Rangachari’s noise tracking method [21]. The last columns show the relative percent EER reduction rates compared to Drugman’s VAD and Rangachari’s method, respectively

From: A robust polynomial regression-based voice activity detector for speaker verification

Noise type SNR level (dB) Proposed algorithm Drugman’s VAD Rangachari’s noise tracking EER reduction compared to Drugman’s EER reduction compared to Rangachari’s
Lynx − 10 36.91 (0.69) 43.80 (0.82) 47.71 (0.88) 15.73 22.63
− 5 27.55 (0.52) 34.51 (0.64) 41.40 (0.78) 20.16 33.45
0 16.67 (0.31) 18.63 (0.34) 28.86 (0.53) 10.52 42.23
5 9.86 (0.18) 10.80 (0.20) 17.76 (0.32) 8.70 44.48
10 6.60 (0.12) 7.61 (0.14) 11.16 (0.20) 13.27 40.86
F16 − 10 42.13 (0.79) 47.50 (0.88) 48.73 (0.89) 11.30 13.54
− 5 33.57 (0.63) 42.78 (0.79) 46.62 (0.86) 21.52 27.99
0 23.71 (0.45) 29.51 (0.55) 37.63 (0.69) 19.65 36.99
5 13.63 (0.25) 15.15 (0.28) 23.93 (0.44) 10.03 43.04
10 8.41 (0.15) 8.77 (0.16) 13.92 (0.26) 4.10 39.58
Car − 10 6.89 (0.12) 5.94 (0.11) 8.70 (0.16) − 15.99 20.80
− 5 5.14 (0.09) 5.57 (0.10) 8.33 (0.15) 7.72 38.29
0 4.78 (0.08) 5.51 (0.10) 8.12 (0.15) 13.24 41.13
5 4.49 (0.08) 5.58 (0.10) 8.04 (0.15) 19.53 44.15
10 4.56 (0.08) 5.58 (0.10) 8.12 (0.14) 18.28 43.82
Babble − 10 37.05 (0.69) 46.33 (0.86) 48.22 (0.89) 20.03 23.16
− 5 27.70 (0.52) 38.50 (0.70) 44.01 (0.81) 28.05 37.06
0 18.05 (0.33) 22.84 (0.42) 33.43 (0.62) 20.97 46
5 11.38 (0.21) 12.25 (0.23) 20.66 (0.38) 7.10 44.91
10 7.25 (0.13) 8.19 (0.15) 12.18 (0.23) 11.47 40.47
Stitel − 10 41.55 (0.78) 45.68 (0.86) 46.84 (0.87) 9.04 11.29
− 5 30.96 (0.58) 36.26 (0.68) 40.24 (0.76) 14.61 23.06
0 19.50 (0.36) 21.32 (0.40) 27 (0.5) 8.53 27.77
5 10.95 (0.20) 11.89 (0.22) 16.24 (0.30) 7.90 32.57
10 6.67 (0.12) 8.48 (0.16) 10.51 (0.20) 21.34 36.53