Ensemble kNN classifiers for human gait recognition based on ground reaction forces

Recognition of people based on their way of movement is one of the most interesting issues of behavioural biometrics. Among the basic characteristics of each biometric system is accuracy. It is currently considered that greater accuracy of biometric can be achieved by ensemble of two and more classifiers. The aim of this study is to present our own method for the usage of ensemble classifiers in the biometrics of the human gait based on ground reaction forces. In the presented ensemble of k-nearest neighbor classifiers the input signals were formed by dividing the ground reaction forces into sub phases characteristic of the support phase of the gait cycle. The research was carried out based on measurements from 200 people (more than 3500 gait cycles). The correct classification rate for proposed here method is more than 97.37%.