SonarGuard2: Ultrasonic Face Liveness Detection Based on Adaptive Doppler Effect Feature Extraction

Xiaoming Zhang, Ke-Yue Zhang, Taiping Yao, Songjun Cao, Shouhong Ding, Long Ma

Abstract: Face anti-spoofing is crucial to security of face recognition systems. Vision-based face liveness detection algorithms often fail when faced with video attacks, such as video replay. However, liveness detection based on sound waves can effectively detect such attacks by relying on the Doppler effect. In order to improve the robustness of liveness detection, we propose a novel framework, named SonarGuard2, adaptively selecting the ultrasonic signals and analyse the Doppler effect. Specifically, we introduce acoustic echo cancellation to filter to get the Doppler effect features, then utilize complex convolutional neural networks to enhance the abilitity to model the Doppler effect features. Furthemore, we propose a novel selection strategy to determine the usability of ultrasonic signals on mobile devices. The performance and visualizaiton on the collected data demonstrate the effectiveness of our framework.

External IDs:dblp:conf/interspeech/ZhangZYCDM25