摘要
提出了基于Kinect传感器深度信息的动态手势识别方法,在预处理阶段通过OpenCV快速跟踪手部,有效分割手势.为改进动态手势轨迹的提取和分类,引入隐马尔可夫模型(HMM)对手势轨迹进行训练和识别.实验结果表明,基于HMM的识别方法对具有时空特性的动态手势有很高的识别率,在不同光照和复杂背景下有鲁棒性的结果.
Based on the depth data obtained from a Microsoft＇s Kinect sensor,this paper proposed a reliable method for dynamic ges- turerecognition. We preprocessed the raw depth datato quickly trackthe hand palm and segment the hand gesture by using OpenCVli- brary. To improve the extraction and classification of the dynamic gesturetrajectory, we introduced the Hidden Markov model for training and recognition. The experimental results show that HMM-based method has high recognition rateand strong robustness in the different illuminationconditions and complex backgrounds.
出处
《厦门大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第4期493-497,共5页
Journal of Xiamen University：Natural Science
基金
国防基础科研计划项目
国防科研重点实验室基金
