J-SPAW: Japanese speaker verification and spoofing attacks recorded in-the-wild dataset
Sayaka Shiota, Suzuka Horie, Kouta Kanno, Shinnosuke Takamichi
In this paper, we present J-SPAW (Japanese speaker verification and spoofing attacks recorded in-the-wild dataset), a novel speech database designed for speaker verification and spoofing detection in-the-wild environments1.J-SPAW is a unique database that simultaneously evaluates speaker verification and spoofing detection under realistic conditions, focusing on physical access scenarios, including replay attacks. The database includes diverse physical access scenarios, enhancing the variety and applicability of datasets available for anti-spoofing research. Our experimental results demonstrate that J-SPAW enables comprehensive analysis of spoofing detection from various perspectives and can be utilized for both speaker verification and spoofing detection tasks. This contribution is expected to advance state-of-the-art speaker verification and spoofing detection and provide a valuable resource for future research.