作者: Ye Tian , Shijun Xiang
DOI: 10.1007/978-3-319-53465-7_2
关键词: Classifier (UML) 、 Spatial analysis 、 Local binary patterns 、 Ordinate 、 Artificial intelligence 、 Replay attack 、 Pattern recognition 、 Discrete cosine transform 、 Computer science 、 Spoofing attack 、 Support vector machine
摘要: Despite the great deal of progress during recent years, face spoofing detection is still a focus attention. In this paper, an effective, simple and time-saving countermeasure against video-based attacks based on LBP (Local Binary Patterns) multiscale DCT (Discrete Cosine Transform) proposed. Adopted as low-level descriptors, features are used to extract spatial information in each selected frame. Next, performed along ordinate axis obtained information. Representing both temporal information, high-level descriptors (LBP-MDCT features) finally fed into SVM (Support Vector Machine) classifier determine whether input video facial attack or valid access. Compared with state art, excellent experimental results proposed method two benchmarking datasets (Replay-Attack CASIA-FASD dataset) have demonstrated its effectiveness.