摘要
为克服步态轮廓变化对步态识别的不利影响,采用步态能量图改进对数Gabor相位一致性特征,提出一种新的步态识别方法。利用局部能量计算方法及频率扩展与噪声补偿策略,使提取的步态特征更具识别性和定位性,并对该步态特征进行线性判别分析降维。应用基于欧氏距离的最近邻分类器在CASIA和USF步态数据库上进行测试,结果表明该方法在个体携包行走、穿着和视角变化的情况下均能较好地识别步态轮廓,相比现有步态识别方法具有更高的正确识别率。
To overcome the adverse effects of gait silhouette change on gait recognition,this paper proposes a new gait recognition method by improving Log-Gabor phase congruency feature of gait energy map. Based on the improved local energy calculation method, frequency expansion and noise compensation strategy, the extracted gait features are more recognizable and localizable. And the dimensions of gait features are dimension reduced by Linear Discriminant Analysis （LDA）. The nearest neighbor classifier based on Euclidean distance is tested on the CASIA and USF gait database. The results show that gait silhouette of individuals can still be well recognized under the influence of walking with package, clothing and view changing, and the correct recognition rate is higher than that of other gait recognition methods.
出处
《计算机工程》
CAS
CSCD
北大核心
2017年第10期198-202,208,共6页
Computer Engineering
基金
国家自然科学基金(61303132)
吉林省教育厅"十三五"基金(吉教科合字[2016]第349号)
关键词
步态能量图
相位一致性
局部能量
频率扩展
噪声补偿
线性判别分析
gait energy map
phase congruency
local energy
frequency expansion
noise compensation
Linear Discriminant Analysis （LDA）
