Speaker Recognition Using Wavelet Cepstral Coefficient, I-Vector, and Cosine Distance Scoring and Its Application for Forensics

Table 4

Average time cost of the speaker recognition models.

Model	Feature extraction (s/speech)		Speaker classification (s/speech)
	Short-term feature extraction	I-vector training	Training	Recognition
MFCCGMM	0.45	—	1.92	0.81
FMFCCPNN	1.29	—	—	0.83
MFCCICDS	0.51	2.21	—	0.85
WCCICDS	2.91	2.19	—	0.89