Robust Digital Speech Watermarking For Online Speaker Recognition

Table 3

The effect of robust watermarking on the performance of speaker verification for different speech databases.

Database	System	EER (%) without robust WM	EER (%) with robust WM
TIMIT	i-vector + MFCC	0.71	4.52
	i-vector + LPRC	1.45	1.59
	GMM-UBM + MFCC	0.79	2.39
	GMM-UBM + LPRC	1.90	5.24
MIT	i-vector + MFCC	15.04	15.34
	i-vector + LPRC	24.20	24.20
	GMM-UBM + MFCC	46	46.02
	GMM-UBM + LPRC	49.43	49.66
MOBIO	i-vector + MFCC	45.96	46.20
	i-vector + LPRC	46	47.80
	GMM-UBM + MFCC	46.66	47.33
	GMM-UBM + LPRC	47	47.59