基于步态能量图特征与稀疏编码的步态识别

Gait Recognition Based on Gait Energy Image Feature and Sparse Coding

为了准确地进行身份识别,提出一种基于步态能量图特征和稀疏编码的步态识别方法。该方法先使用训练数据生成识别字典及其原子标签信息,再将待识别视频中提取的步态能量图SIFT特征使用识别字典进行稀疏表示,最后通过投票的方式得到识别结果。实验数据选用中科院步态数据库CASIA,实验结果表明,所提算法识别率高,计算速度快。

Proposes a gait recognition method based on Gait Energy Image(GEI)features and dictionary learning to identify more accurately.The method utilizes the training data to generate the recognition dictionary and its atomic tag information,and then uses the recognition dictionary to represent the GEI SIFT feature extracted from the video to be identified.Finally,the recognition result is obtained by a voting operation.The experimental results on dataset CASIA B improve that the proposed method has high recognition rate and fast computation speed.