Abstract:
Being the very first in the category of low-cost consumer-level depth sensors, Microsoft Kinect has opened the door to a new generation of computer vi- sion and biometric security applications after its release. This thesis focuses on designing new methodologies for Kinect-based gait recognition systems that utilize the Kinect 3D virtual skeleton to construct effective and robust mo- tion representations. Our goal is to propose a gait recognition method that focuses on designing a feature descriptor that can capture person-specific dis- tinct motion patterns, caused by the influence of human physiology and be- havioral traits. In this regard we use pre-existing representations of skeletal data namely Joint Relative Distance and Joint Relative Angle. The proposed methodologies contain more representations using mean and standard devi- ation of the data which can effectively handle view and pose variations. We used a dynamic time warping-based kernel that takes a collection of sequences as parameters and computes a dissimilarity measure between the training and the unknown sample. It can effectively handle variable walking speed without any need of extra pre-processing. The effectiveness of the proposed method- ologies are evaluated using 3D skeletal gait database captured with a Kinect v1 sensor. In our experiments, fusion of mean and standard deviation achieves promising results, as compared against the previous implementations.
