A Multiscale Chaotic Feature Extraction Method for Speaker Recognition

Table 6

The average EER value in all experiments.

Feature	Speech type	Acoustic feature	Nonlinear feature	Chaotic feature	i-vector model	Compare with i-vector
EER (%)	Clean	2.562	2.833	2.149	2.497	↓13.94%
	White	14.96	15.68	10.74	14.82	↓26.5%
	Babble	13.60	14.43	8.34	11.14