Optimizing resources on smartphone gait recognition

Abstract

Inertial gait recognition is a biometric modality with increasing interest. Gait recognition in smartphones could become one of the most user-friendly recognition systems. Some state-of-art algorithms need to perform cross-comparisons of gait cycles to obtain a comparison result. In this contribution, two facts are studied in order to reduce the computational cost: the influence of using representative gait cycles and the gait signals length. The results obtained show that cross-comparisons could be performed with representative gait cycles without heavily penalizing accuracy and reducing computational cost, and that selecting representative gait cycles from the end of the signal perform better that the ones on the beginning.