Kinect gait skeletal joint feature-based person identification

摘要

Gait not only defines the way a person walks, but also provides interesting cues on individuals daily routine, mental state, health condition or even cognitive function. The importance of incorporating cognitive behavior and analysis in biometric systems has been noted recently. In this article, we develop a biometric-security system using gait-based skeletal information from Microsoft Kinect v1 sensor. The gait cycle is calculated by detecting the three consecutive local minima between the distance of left and right ankle joints. We have utilized the distance feature vector for each of the joints with respect to other joints in the gait cycle for extraction. Mean and variance features are extracted from the distance feature vector. The K Nearest Neighbors (KNN) algorithm is used for classification purpose. The classification accuracy of our proposed approach is 93.33%. The effectiveness of the method is evaluated by comparing it with others existing approaches. Experimental results show that proposed approach is having better recognition accuracy compared to other approaches. Incorporating this biometric in situation awareness system that can identify the mental state of a human is the future direction of this research.