作者: R. Vera-Rodriguez , J. S. D. Mason , J. Fierrez , J. Ortega-Garcia
DOI: 10.1007/978-3-642-17289-2_47
关键词: Pattern recognition 、 Focus (optics) 、 Data mining 、 Range (statistics) 、 Computer science 、 Biometrics 、 Reference model 、 Feature (machine learning) 、 Artificial intelligence 、 Support vector machine 、 Ground reaction force 、 Time domain
摘要: This paper reports an experimental analysis of footsteps as a biometric. The focus here is on information extracted from the time domain signals collected array piezoelectric sensors. Results are related to largest footstep database date, with almost 20,000 valid and more than 120 persons, which well beyond previous databases. Three feature approaches have been extracted, popular ground reaction force (GRF), spatial average upper lower contours pressure signals. Experimental work based verification mode holistic approach PCA SVM, achieving results in range 5 15% EER depending conditions quantity data used reference models.