Leveraging shape and depth in user authentication from in-air hand gestures

摘要: Depth-sensors, such as the Kinect, have predominately been used a gesture recognition device. Recent works, however, proposed using these sensors for user authentication biometric modalities as: face, speech, gait and gesture. The last of — gestures, in context full-body hand-based is relatively new but has shown promising performance. In this paper, we focus on gestures that are performed in-air. We present novel approach to from by leveraging temporal hierarchy depth-aware silhouette covariances. Further, investigate usefulness shape depth information modality, well importance hand movement when performing By exploiting both our method attains an average 1.92% Equal Error Rate (EER) dataset 21 users across 4 predefined hand-gestures. Our consistently outperforms related methods dataset.