复杂背景下人体步态分析方法

Analysis on Human Gait Characteristics Under Complex Background

为了获得复杂背景下的人体步态信息,提出了基于超像素分割的步态提取方法,通过对关键帧步态能量图构造Gabor二维滤波器对步态能量图进行卷积提取特征,构造5个尺度,8个方向共40个Gabor滤波器组与步态能量图进行卷积操作特取特征,为了更有效地表征特征数据,使用主成分分析进行特征降维,最后选取无监督的分类方法-K均值聚类法进行分类识别,实验结果表明,该方法能够实现步态的有效识别。

In order to obtain human gait information under complex background,a method of gait extraction based on super-pixel segmentation is proposed.By constructing Gabor two-dimensional filter for key frame gait energy map,the features of gait energy map are extracted by convolution,and the convolution operation of 40 Gabor filter groups and gait energy map in 8 directions and 5 scales is constructed.In order to represent the feature number more effectively,according to the research,principal component analysis is used to reduce the dimension,and finally unsupervised classification method K-means clustering method is selected for classification and recognition.The experimental results show that this method can achieve effective gait recognition.