研究目的
Investigating the capability of temporal deep neural networks to interpret natural human kinematics for active biometric authentication with mobile inertial sensors.
研究成果
The study confirms that natural human kinematics convey necessary information about person identity and can be useful for user authentication on mobile devices. The proposed Dense Clockwork RNN can be successfully applied to other tasks based on analysis of sequential data, such as gesture recognition from visual input.
研究不足
The study is limited by the technical constraints of mobile devices, including computational power for model adaptation to a new user and for real-time inference, as well as the absence (or very limited amount) of 'negative' samples.
1：Experimental Design and Method Selection:
The study compares several neural architectures for efficient learning of temporal multi-modal data representations, including an optimized shift-invariant dense convolutional mechanism.
2：Sample Selection and Data Sources:
A dataset of human movements was passively collected by 1500 volunteers using their smartphones daily over several months.
3：List of Experimental Equipment and Materials:
Smartphones with inertial sensors (accelerometer, gyroscope, and magnetometer) were used for data collection.
4：Experimental Procedures and Operational Workflow:
Data was recorded from the moment after the phone was unlocked until the end of a session, with accelerometer and gyroscope sensors sampled at 200 Hz and magnetometer at 5 Hz.
5：Data Analysis Methods:
The study incorporates discriminatively trained dynamic features in a probabilistic generative framework, analyzing data with PCA-reduced feature space and GMMs with 256 mixture components.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
