Statistical modeling for human gait analysis based on field data

本文研究了基于视频数据的步态配准和建模方法，对正视图中的人体步态进行分析和分类。首先建立了恒速统计步行模型，并利用尺度变化参数和表征人类步行速度的速度参数成功地估计了步态模型中包含的参数。我们还研究了模型和估计精度的运动捕捉系统，给我们精确的测量运动物体的轨迹。我们的基本统计步态模型只包括两个参数的注册和速度，我们可以很容易地估计参数从正面视图步态视频数据。 其次，我们扩展了这个简单的模型，以处理速度变化和其他人体运动参数。柯林斯等人（2009）已经报道了手臂摆动在基于简单步态模型的步态运动中是非常重要的作用。我们考虑了基于柯林斯等人（2009）的人类步态建模。

In this study we studied the problem of analyzing and classifying frontal view human gait by registration and modeling on a video data. Firstly we created the statistical walking model with constant speed and succeeded in estimating parameters included in the gait models with the scale changing parameter and the speed parameter that indicates the speed of human walking. We also investigated the model and estimation accuracy with the motion capture system which gave us precise measurements of trajectories of moving objects. Our basic statistical gait model only includes two parameters of registration and speed and we can easily estimate the parameters from frontal view gait video data. Secondly, we extended this simple model to handle the speed changing and other human movement parameters. Collins et al. (2009) has reported that arm swing is an very important role in the gait motion based on the simple gait model. We considered the human gait modeling based Collins et al. (2009).