摘要
随着低成本深度传感器(如微软Kinect)的出现,人体行为识别研究吸引了很多研究人员.由于这些设备提供了身体关节的三维位置等骨骼数据,使得基于骨骼的人体行为识别变得简单.但这些关节特征的信息存在部分冗余或者不必要的肢体特征,从而降低识别精度.为此,提出一种智能算法来优化关节点信息的方法过滤掉一些不必要的关节点的特征信息,从而提高识别精度.实验结果表明,提出的方法在UTKinect数据集上测试得到的精度达到97.39%,在Florence3D数据集上测试得到的精度达到93.05%.
With the low-cost depth sensors （such as the Microsoft Kinect） developed, human action recognition has attracted lots of researchers. Since these devices provided skeletal data consisting of 3D posi- tions of body joints, human action recognition became simple. But these might contain irrelevant or redundant features information of body joints that could cut down recognition accuracy. A genetic algorithm is used to optimize features information of body joints with filtering unnecessary data and raise recognition accuracy. The proposed approach has tested on UTKinect-Action dataset and Florence3D-Action dataset, experimental results show that our method gains test accuracy with 97.39% and 93.05% respectively.
出处
《广西民族大学学报（自然科学版）》
CAS
2017年第1期90-94,共5页
Journal of Guangxi Minzu University ：Natural Science Edition
基金
广西自然科学基金(2015GXNSFAA139311)
关键词
行为识别
特征选择
遗传算法
action recognition
features selection
Genetic Algorithm
