Combining Homogeneous Base Classifiers to Improve the Accuracy of Biometric Systems Based on Ground Reaction Forces

In recent times, one of the most interesting and fast-growing issues is biometrics. Due to its unique properties, human gait recognition deserves special attention. Among more desirable properties of biometric systems is the accuracy of recognition. Higher quality of the operation of the biometric system, understood as a higher correct classification rate, can be obtained with the use of combined classifiers. The purpose of this paper is to present the results of operation of the three biometric systems, in which different strategies were used to connect homogeneous base classifiers. The measured ground reaction forces were divided into subphases. To determine the distance between the measured and the model signals, dynamic time warping algorithm was used. The experiments were carried out on a group of 220 people (more than 4000 gait cycles). The results (up to 98.95% for 220 subjects) of correct diagnoses are very good and surpass the results of other biometric systems operating on signals derived from floor sensors.