Conclusion and Remarks

Abstract

The study of behavioral biometrics has revealed a number of powerful new person-distinguishing characteristics, some of which have the potential to be less intrusive and more fraud-resistant when compared to other security mechanisms such as physical biometrics. However, the complicated data samples that are often associated with the behavioral biometrics impose the need for devising a new generation of processing and classification techniques; that is, techniques that are able to identify the key traits while not being significantly influenced by the intra-person variability. In this book, a comprehensive analysis of such techniques has been presented along with the sources of variability, with the intended purpose of performing person recognition via the behavioral biometric known as footstep ground reaction force (GRF), which is a form of gait biometric. Through our work, two novel machine learning-based normalization techniques were proposed, which support two assertions related to the effects of shoe type and stepping speed on the GRF recognition performance. Furthermore, we have compared a number of feature extractor, normalizer, and classifier configurations that had never before been cross-examined with respect to GRF-based gait recognition. In this concluding chapter, we summarize these findings and we focus on ways in which the work presented in this book may be improved upon by the future research. The trends and challenges that we believe will shape the future development of gait biometrics are also explored. Finally, we discuss some of the most recent advances in gait biometric recognition and where they may fit with respect to the work presented in this book