Skip to main content

Advertisement

Table 5 Male speaker verification results of i-vector 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 30.04 (0.55) 43.57 (0.81) 43.57 (0.81) 31.05 31.05
− 5 17.58 (0.33) 28.36 (0.53) 29.51 (0.55) 38.01 40.42
0 9.48 (0.17) 14.37 (0.27) 17.43 (0.32) 34.03 45.61
5 5.58 (0.09) 8.25 (0.14) 11.39 (0.21) 32.36 51
10 3.90 (0.06) 5.35 (0.09) 7.72 (0.14) 27.10 49.48
F16 − 10 38.60 (0.73) 48.16 (0.89) 48.93(0.89) 19.85 21.11
− 5 27.44 (0.52) 38.07 (0.71) 37.08 (0.70) 27.92 26
0 15.82 (0.30) 21.71 (0.40) 23.39 (0.43) 27.13 32.36
5 8.48 (0.15) 11.31 (0.21) 14.98 (0.27) 25.02 43.39
10 5.35 (0.09) 7.41 (0.13) 10.16 (0.18) 27.8 47.34
Car − 10 3.74 (0.06) 4.66 (0.08) 6.04 (0.11) 19.74 38.08
− 5 3.28 (0.05) 4.43 (0.07) 3.66 (0.06) 25.95 10.38
0 2.98 (0.05) 4.20 (0.06) 5.58 (0.09) 29.04 46.59
5 3.13 (0.05) 4.05 (0.06) 5.65 (0.09) 22.71 44.60
10 3.13 (0.04) 3.97 (0.06) 5.58 (0.09) 21.15 43.90
Babble − 10 31.88 (0.60) 47.24 (0.87) 47.09 (0.88) 32.51 32.3
− 5 19.95 (0.37) 32.95 (0.61) 42.66 (0.80) 39.45 53.23
0 10.85 (0.19) 18.19 (0.34) 20.18 (0.38) 40.35 46.23
5 5.65 (0.10) 9.25 (0.17) 12.00 (0.22) 38.92 52.91
10 4.35 (0.07) 6.11 (0.11) 8.56 (0.15) 28.80 49.18
Stitel − 10 37.53 (0.71) 46.56 (0.87) 45.87 (0.86) 19.39 18.18
− 5 22.24 (0.42) 32.11 (0.60) 31.72 (0.60) 30.73 29.88
0 11.23 (0.20) 17.35 (0.32) 15.75 (0.29) 35.27 28.7
5 5.81 (0.11) 9.17 (0.17) 10.24 (0.19) 36.64 43.26
10 4.05 (0.07) 6.34 (0.11) 7.41 (0.13) 36.12 45.34