Abstract: Gait template plays a crucial role in improving real-time performance of gait recognition. However, its performance suffers from lacking of time-series information and insufficient extraction of statistical features in images. In this paper, Gait Energy Image（GEI） and the Chrono-Gait Image（CGI） is used for the representation of gait features, followed by extracting their respective spatial features and further extracting spatial feature. Based on these features, an integrated HOG gait template is proposed. This template can preserve time-series information well and effectively represent spatial features in gait sequences. Experiments in USF benchmark gait database show that compared with other published algorithms, the proposed template achieves better performance in gait recognition.

Key words: computer vision, gait recognition, gait template, Histograms of Oriented Gradients（HOG） feature

摘要： 步态模板在提升步态识别的实时性能中扮演了关键角色。由于缺乏时间信息和不能充分提取步态中的统计特征，其识别性能会受到一定的损害。以步态能量图（GEI）为模板，并使用基于时间保持的步态能量图（CGI），从这两个模板中进一步提取空间特征。在此基础上，构造了集成HOG步态模板。这一模板能较好地保持时间信息和有效地提取空间结构特征。在USF步态数据集的实验表明，与其他已知步态识别方法相比，提出的模板实现了好的识别性能。

关键词: 计算机视觉, 步态识别, 步态模板, 方向梯度直方图（HOG）特征