基于DTW的Kinect动作识别方法研究与实现
首发时间:2013-07-19
摘要:虚拟现实技术为基于三维的产品展示提供了全新的人机交互方式。为提高虚拟三维展示系统的人机交互性,使用Kinect传感器获取场景三维景深图和人体骨骼图。把语音识别领域中的经典动态时间归整(Dynamic Time Warping,DTW)算法应用到Kinect动作识别中。通过Kinect获取人手相应动作点三维坐标,并取一段时间窗内动作点数组与模板库中样本比对,计算最短路径;再经过识别系统与专家知识进行匹配,识别动作。通过实际展示系统验证了算法的有效性。
关键词: 通信与信息系统 虚拟现实 Kinect 动态时间归整算法 动作识别
For information in English, please click here
Research and implementation of Kinect action recognition method based on DTW
Abstract:The new human-computer interaction mechanism of three-dimensional production display is enhanced by virtual reality technology. In order to improve the human-computer performance of virtual reality production system, the Kinect sensor is used to get the three-dimensional scene depth and human skeleton map. The classical Dynamic Time Warping (DTW) algorithm is applied to the Kinect action recognition. The three dimensional action coordinate is obtained by Kinect and compared with the samples in template library to calculate the shortest path. After matching based on expert knowledge, the action is identified. The algorithm effectiveness is verified by the actual equipment display system.
Keywords: communication and information system virtual reality Kinect dynamic time warping algorithm action recognition
基金:
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于DTW的Kinect动作识别方法研究与实现
评论
全部评论