A study on gait parameter estimation stability for the frontal view gait video data based on simulation

We study the problem of analyzing and classifying frontal view gait video data. The video data filmed from the frontal view is difficult to analyze, because the subject getting close in to the camera, and data includes the scale-changing parameters (Barnich and Droogenbroeck 2009, Lee et al. 2008). To cope with this, Okusa et al. (2011) and Okusa & Kamakura (2012) proposed a registration for scales of moving object using the method of nonlinear least squares, but Okusa et al. (2011) and Okusa & Kamakura (2012) did not focus on the human leg swing. Okusa & Kamakura (2013c) focus on the gait analysis using arm and leg swing model with estimated parameters and application to the normal/ abnormal gait analysis. However, their models have many of parameters, and it raise calculation cost and instability of parameter estimation. Okusa & Kamakura (2013a) focus on the calculation cost and parameter estimation stability. The performance of Okusa & Kamakura (2013a) model is able to speed up the parameter estimation. However, the problem of parameter estimation stability still remains to be solved. Okusa & Kamakura (2013b) proposed simplified gait model based on the Okusa & Kamakura (2013a)'s model, it settled stability of parameter estimation. In this article, we focus on the behavior of Okusa & Kamakura (2013b) model's parameters. We validate the Okusa & Kamakura (2013b)'s model from the stand point of stability of the parameter estimation based on the numerical simulation. As a result, Okusa & Kamakura (2013b)'s gait model is stable to estimate the arm swim amplitude, subjects walking speed, gait frequency. However, on the other hand, this model is difficult to estimate the phase parameters. This result indicates Okusa & Kamakura (2013b) model is difficult to apply for the frontal view gait authentication.