Skip to Main Content
A novel biometric recognition system is designed in this paper, using ground reaction force (GRF) measurements of continuous gait. Original GRF signals are combined in three directions Respectively. Waveform interpolation and align ment are performed to meet th e demand of feature extraction ba sed on wavelet packet (WP) decomposition, re-sampling approach is utilized to expand valid training sets. Features are s elected u sing a fuzzy set-based features election criterion. Classification is accomplished using a kernel-based support vector machine (SVM). The parameter tuning of the S VM classifier is performed using a grid searching method. The approach is tested on a database comprising GRF records obtained from 103 subjects. Comparative results demonstrate that re-sampling approach and waveform interpolation and alignment can improve the recognition accuracy.