查询结果:   吴晓雨,杨成,冯琦.基于Kinect的手势识别算法研究及应用[J].计算机应用与软件,2015,32(7):173 - 176,276.
中文标题
基于Kinect的手势识别算法研究及应用
发表栏目
人工智能与识别
摘要点击数
1080
英文标题
RESEARCH ON KINECT-BASED HAND GESTURE RECOGNITION ALGORITHM AND ITS APPLICATIONS
作 者
吴晓雨 杨成 冯琦 Wu Xiaoyu Yang Cheng Feng Qi
作者单位
中国传媒大学信息工程学院 北京 100024     
英文单位
College of Information Engineering,Communication University of China,Beijing 100024,China     
关键词
手势识别 人机交互 Kinect传感器 数字电视的模拟控制
Keywords
Hand gesture recognition Human computer interaction Kinect sensor Digital TV simulation control
基金项目
国家科技支撑课题(2012BAH01F00)
作者资料
吴晓雨,讲师,主研领域:图像处理,人机交互技术。杨成,副教授。冯琦,硕士生。 。
文章摘要
手势识别技术是人机交互技术的重要研究内容。为了提高基于Kinect的手势识别性能,提出基于深度人手定位和hog特征的静态手势识别算法及基于改进HMMs的动态手势识别算法。静态手势识别算法首先通过Kinect的深度信息完成人手定位,而后在定位区域内提取基于梯度方向直方图的形状特征并利用级联Adaboost训练的手势模型,实现对静态手势的准确识别,在公开手势数据库中测试的实验结果表明提出的静态手势识别算法具有较高的识别率。动态手势识别算法首先通过Kinect获取手心轨?⑻崛」旒G邢呓嵌茸魑卣鳎酶慕囊矶品蚰P褪迪侄质频呐斜穑笛榻峁砻魈岢龅亩质剖侗鹚惴ㄏ啾扔诖矵MMs算法有效地排除了无效手势。此外利用提出的动静态手势识别算法有效地控制了模拟的数字电视。
Abstract
Hand gesture recognition technique is the important research content in human-computer interaction technology. In order to improve the performance of Kinect-based hand gesture recognition, we propose the static hand gestures recognition algorithm, which is based on depth hands localisation and hog feature, and the dynamic hand gesture recognition algorithm which is based on the improved HMMs. For the static hand gestures recognition algorithm, first it achieves the hand localisation through Kinect depth information, and then extracts the gradient orientation histogram-based shape features within localisation area and uses the gesture model trained by Adaboost cascade to realise the accurate recognition on static gestures. It is demonstrated by the experimental results tested on public hand gesture dataset that the proposed static hand gestures recognition algorithm obtains higher recognition rate. The dynamic gestures recognition algorithm first obtains the palm trajectories using Kinect and extracts the tangent-angle of trajectory as the feature, then employs the improved hidden Markov model to implement the discrimination of hand gestures. Experimental results show that the proposed dynamic hand gestures recognition algorithm effectively excludes the undefined gestures compared with traditional HMMs method. Besides, the proposed dynamic and static hand gestures recognition algorithms are applied in effective control of simulated digital TV.
下载PDF全文