Real time accelerometer-based gait recognition using adaptive windowed wavelet transforms

This paper presents a real time gait recognition system using the wavelet transform. The activity signal is acquired from three-axis accelerometers on mobile phones. It is first decomposed into wavelet coefficients with eight levels. Several statistical measures, such as power, mean, variance, energy, and the energy of neighbor difference, are calculated from these coefficients. Furthermore, the adaptive window size is adopted to well fit the footstep of each person. The selected features are also adjusted adaptively to improve the accuracy. The simulation results show that the proposed method has reliable recognition accuracy both in the real-time and the long-term cases.