步态特征精密提取与身份认证(英文)

Precision Extraction of Gait Feature and Identity Authentication

为更有效地检测人体目标,弥补单一模型在步态特征提取中的不足,提出了基于双模型的步态特征精密提取方法,且构建了基于步态特征进行身份认证的门禁监控实验平台.首先从摄像机捕获步态视频输入计算机,发现人体目标后对其进行检测与跟踪;然后分割人体轮廓并将其规格化叠加处理获取步态特征图;为精确提取步态特征,将人体整体模型与简化模型相结合,提取步态参数作为识别参量输入支持向量机(SVM)进行分类识别,正确识别率(PCR)为77%～80%.结果表明该方法有助于步态特征的精密提取,且实验平台能较好地自动监控人体目标并进行身份认证.

To better detect human body and remedy defects of feature extraction with a single model, a method for precision extraction of gait feature based on double model was proposed to obtain gait parameters, and an access monitoring platform for identity authentication was presented. Firstly, the gait video captured by camera was input to the computer to detect target and monitor the access. Secondly, body silhouettes were extracted and normalized to obtain the gait feature image. Thirdly, the integral model was combined with the simplified model to extract gait parameters. Finally, the technique of support vector machines was presented for identity authentication. The probability of correct recognition （PCR） has achieved 77% --80%. This method based on integral model and simplified model is helpful for precision extraction of gait feature. The platform can automatically detect human body and provide a better way for identity authentication.