Feature Extraction of Gait Data via PCA and Personal Identification Using Neural Networks

Gait analysis is used in a wide range of fields, including clinical applications such as diagnosis and rehabilitation, investigation of gender and racial differences, and personal identification, which is important in forensic investigations and security. In recent years, gait data can be obtained from force plates, 3D motion capture, RGB cameras, and so on. In this study, we applied principal component analysis (PCA) to gait data captured with an RGB camera and analyzed the main data characteristics. However, the data used here is joint information whose posture is estimated using neural networks (NN) from gait videos taken with the camera. Furthermore, we created a NN model for personal identification using data whose dimensionality was reduced using PCA and attempted the identification.