作者: Dipjyoti Paul , Monisankha Pal , Goutam Saha
DOI: 10.1109/INDICON.2015.7443805
关键词: Spoofing attack 、 Pattern recognition 、 Voice activity detection 、 Biometrics 、 Artificial intelligence 、 Speech synthesis 、 Hidden Markov model 、 Speech recognition 、 Feature extraction 、 Block (data storage) 、 Mel-frequency cepstrum 、 Computer science
摘要: Now-a-days, speech-based biometric systems such as automatic speaker verification (ASV) are highly prone to spoofing attacks by an imposture. With recent development in various voice conversion (VC) and speech synthesis (SS) algorithms, these can pose a serious potential threat the current state-of-the-art ASV systems. To impede enhance security of systems, efficient anti-spoofing algorithms is essential that differentiate synthetic or converted from natural human speech. In this paper, we propose set novel features for detecting attacks. The proposed computed using alternative frequency-warping technique formant-specific block transformation filter bank log energies. We have evaluated existing against several kinds data ASVspoof 2015 corpora. results show techniques outperform approaches attack detection task. investigated paper also accurately classify equal error rates (EERs) 0% been achieved.