Paper Info
Title
Replay Attacks Detection Using Phase and Magnitude Features with Various Frequency Resolutions
Abstract
Replay attacks detection is a challenging task under unseen recording environments and playback devices. To overcome this problem, we propose a combination of phase-based and magnitude-based feature with various frequency resolutions. Conventional feature used in replay attacks detection does not pay much attention to phase which is expected to be effective for replay attacks detection because the genuine and spoof speech has significant differences in phase. In this paper, we investigate the combination of phase-based and magnitude-based features with different frequency resolutions to better solve the classification issue. The replay attacks detection results on ASVspoof 2017 challenge indicated that our proposed approach achieved 51.3% relative error reduction rate than the conventional magnitude-based feature.
Year
DOI
Venue
2018
10.1109/ISCSLP.2018.8706628
2018 11th International Symposium on Chinese Spoken Language Processing (ISCSLP)
Keywords
Field
DocType
Feature extraction,High frequency,Mel frequency cepstral coefficient,Task analysis,Time-frequency analysis,Phase frequency detector
Magnitude (mathematics),Mel-frequency cepstrum,Task analysis,Pattern recognition,Computer science,Speech recognition,Feature extraction,Time–frequency analysis,Artificial intelligence,Replay attack,Approximation error,Phase frequency detector
Conference
ISBN
Citations 
PageRank 
978-1-5386-5627-3
0
0.34
References 
Authors
0
6
Name
Order
Citations
PageRank
Meng Liu13918.70
Longbiao Wang227244.38
Zeyan Oo351.49
Jianwu Dang429391.90
Dongbo Li5314.92
Seiichi Nakagawa6598104.03