Abstract: Walking contains body shape and dynamic information，most gait recognition algorithms are based on single one，and can’t get satisfied recognition rate.According to the idea of feature fusion，a gait recognition algorithm based on the fusion of body contour and angles of lower limbs is proposed.Body contour is described by using Fourier-based descriptors.Different from the model-based methods for getting dynamic information，the knowledge in body anatomy is used for getting the angles of lower limbs，so computational cost is low.Weighted fusion method is used for realizing features fusion.Experimental results demonstrate that the performance of proposed algorithm is much better than those based on single feature.

Key words: biometrics, gait recognition, feature fusion, Fourier-based descriptors

摘要： 步态运动中包含人体形状信息和运动信息，目前步态识别算法多数基于单一信息，不能取得满意的识别结果。利用特征融合的思想，提出一种融合人体轮廓特征和下肢角度特征的步态识别算法。采用傅立叶描述子描述人体轮廓特征；区别于基于模型的运动特征提取方法，依据人体解剖学的知识获取下肢角度，计算代价较小；采用加权融合规则实现两类特征的融合。仿真结果表明，本算法的性能较基于单个特征的算法有明显的提高。

关键词: 生物特征识别, 步态识别, 特征融合, 傅立叶描述子