Identification of Weakly Pitch-Shifted Voice Based on Convolutional Neural Network

Table 3

Detection performance of weakly pitch-shifted voice in binary classification.

Pitch shifting software	Training dataset	Testing dataset	Detecting method
			[6] LFCC + GMM		[8] MFCC + GMM		Proposed
			Rate	FAR	Rate	FAR	Rate	FAR
Audition	TIMIT	TIMIT	98.11	0.83	97.29	1.34	98.72	0.70
	TIMIT	UME	92.95	5.50	93.25	1.67	96.83	1.84
	UME	TIMIT	96.72	0.47	95.21	1.72	97.26	0.52
	UME	UME	97.70	0.88	97.82	0.64	96.82	0.91
GoldWave	TIMIT	TIMIT	97.92	0.68	98.93	0.42	98.14	1.47
	TIMIT	UME	82.86	14.60	91.56	4.64	92.98	5.95
	UME	TIMIT	92.58	0.13	93.93	0.25	96.84	1.25
	UME	UME	98.39	0.08	98.78	0.14	97.79	0.92
Audacity	TIMIT	TIMIT	98.27	0.32	99.55	0.06	99.10	0.29
	TIMIT	UME	83.04	15.44	87.96	10.07	94.25	4.05
	UME	TIMIT	91.89	0.06	91.84	0.03	98.12	0.33
	UME	UME	98.89	0.09	99.30	0.09	98.39	0.87