User Authentication Method via Speaker Recognition and Speech Synthesis Detection

Table 1

The dataset used in the evaluation.

Voice authentication module		Synthetic speech detection module
Training data (MFCC)	Testing data (MFCC)	Training data (Mel-Spectrogram)	Testing data (Mel-Spectrogram)
USR 1 ∼ 8	USR1 ∼ 8	USR1 ∼ 8 (Natural & Synthetic)	User1 ∼ 5 (Natural), User6 ∼ 8 (Synthetic)
The number of MFCC samples		The number of Mel-Spectrogram samples
44,758	9,791	27,974	6,993