Skip to main content

Advertisement

Table 6 Female 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 31.32 (0.58) 41.84 (0.78) 44.52 (0.84) 25.14 29.65
− 5 20.16 (0.38) 29.44 (0.55) 36.76 (0.86) 31.52 45.15
0 11.82 (0.22) 15.80 (0.30) 23.93 (0.44) 25.19 50.60
5 6.81 (0.12) 8.34 (0.15) 14.35 (0.26) 18.34 52.54
10 4.13 (0.07) 4.85 (0.08) 9.64 (0.18) 14.84 57.15
F16 − 10 38.79 (0.71) 46.26 (0.85) 47.71 (0.88) 16.14 18.69
− 5 27.99 (0.52) 37.63 (0.70) 42.20 (0.78) 25.61 33.67
0 17.4 (0.33) 24.43 (0.46) 31.54 (0.59) 28.77 44.83
5 9.93 (0.18) 11.89 (0.22) 19.29 (0.36) 16.48 48.52
10 5.87 (0.10) 6.16 (0.11) 12.54 (0.23) 4.70 53.19
Car − 10 3.62 (0.06) 3.77 (0.06) 6.74 (0.12) 3.97 47.29
− 5 2.82 (0.05) 3.19 (0.05) 6.23 (0.11) 11.59 54.73
0 2.75 (0.04) 3.12 (0.05) 6.09 (0.11) 11.86 54.84
5 2.75 (0.04) 3.04 (0.05) 6.02 (0.11) 9.54 54.32
10 2.75 (0.04) 3.04 (0.05) 6.09 (0.11) 9.54 54.82
Babble − 10 33.21 (0.63) 44.81 (0.84) 46.12 (0.84) 25.88 27.99
− 5 21.68 (0.40) 34.15 (0.63) 40.32 (0.75) 36.51 46.23
0 12.98 (0.24) 20.08 (0.37) 27.19 (0.51) 35.35 52.26
5 6.89 (0.12) 9.42 (0.17) 16.75 (0.31) 26.85 58.86
10 4.06 (0.07) 5.07 (0.09) 10.73 (0.19) 19.92 62.16
Stitel − 10 33.21 (0.63) 46.04 (0.85) 45.17 (0.83) 27.86 26.47
− 5 26.83 (0.50) 34.37 (0.64) 35.53 (0.65) 21.93 24.48
0 15.08 (0.28) 18.92 (0.35) 22.33 (0.42) 20.29 32.46
5 8.12 (0.14) 10.37 (0.19) 13.77 (0.26) 21.69 41.03
10 4.20 (0.07) 6.23 (0.11) 8.99 (0.17) 32.58 53.28