基于步态特征的身份识别算法研究

摘要

步态识别根据人走路的姿势进行身份识别,由于人在行走时在空间呈现出的不同几何模式,单一几何特征难以全面描述步态特征,导致身份识别正确率不高.为提高身份识别的正确率,提出一种空间和频率特征模式相融合的身份识别算法.首先利用摄像机采集步态图像序列,然后分别采用极坐系和傅里叶变换提取步态空间特征和频率特征,并对两种特征进行融合,最后采用支持向量机对融合特征进行学习和分类,进行身份识别.结果表明,相对于单一步态特征为参数的身份识别算法,融合算法的身份识别正确率有了明显提高,且具有更好的稳定性.%Gait image contains spatial and frequency characteristics. The current recognition algorithms use the space character or frequency characteristics as the syndrome parameters of identification, lead to low correct rate of I-dentity recognition. The paper put forward an identification algorithm based on gait spatial characteristics and frequency feature fusion. First, it used camera to acquire gait image sequences, and then used polar sit and Fourier -transform to extract gait spatial characteristics and frequency characteristics. Finally, support vector machine was used to leam and classify the fusion characteristics. The results show that, compared with the single gait characteristics identification algorithm, the fusion algorithms has improved the correct rate of identification, and has a better stability.