One-Class Fake Speech Detection Based on Improved Support Vector Data Description

Table 1

Summary of the ASVspoof 2019 LA and “In-the-wild” datasets.

	Bonafide		Spoofed
	#Speaker	#Utterance	#Speaker	#Utterance
ASVspoof 2019
Training	20	2,580	20	22,800
Development	20	2,548	20	22,296
Evaluation	68	7,533	68	63,882
In-the-wild
Evaluation	54	19,963	54	11,816
FoR
Training	—	26,941	33	26,927
Development	—	5,400	33	5,398
Evaluation	—	2,264	33	2,370