文档介绍：: .
Hans汉斯
Instrumentation and Equipments 仪器与设备'2020, 8(1), 8-20
Published Online March 2020 in Hans, https://doi・org/10・12677/iae・2020・81002
KinectV2 Sensor Real-Time Dynamic Gesture Recognition Algorithm
Xidong Liu1, Yan Meng1, Guoyou Li2, Lei Tang2, Jiadong Guo2
1College of Biological and Environmentai Engineering, Tianjin Vocational Institute University, Tianjing 2Shanghai Institute of Aerospace Technology, Shanghai
Email: *************@
Received: Feb. 15thz 2020; accepted: Mar. 3巴 2020; published: Mar. 10th, 2020
Abstract
In order to solve the problems of low generalization of features extracted by the dynamic gesture ac・ tion recognition algorithm, poor real-time recognition and low recognition rate, this paper proposes a template matching gesture recognition method based on KinectV2 sensor. This method uses Kinect Studio and Visual Gesture Builder to build a gesture action database, extracts the angle between key bones of the human body as a one-dimensional feature vector that changes with time, and proposes Weighted DTW gesture motion recognition algorithm based on the dynamic time warping (DTW) algorithm with the help of dynamic planning ideas・ Through the improvement of the DTW algorithm, the regular path converges faster, the real-time and accuracy of gesture motion recognition is im・ proved, and the real-time and validity verification experiments are carried out in the Unity 3D virtual simulation experiment platform for simple human-computer interaction・
Keywords
Kinectv2 Sensor, Gesture Recognition, Skeletal Angle Characteristics, Human Computer Interaction
KinectV2传感器实时动态手势识别算法
刘希东打孟岩打李国友打唐磊訂郭佳栋彳
I天津职业大学生物与环境工程学院，天津
2上海航天技术研究院，上海
Email: *************@
收稿日期：2020年2月150：录用日期：2020年3月3日：发布日期：2020年3月10日
为解决动态手势动作识别算法提取的特征泛化程度低，识别实时性差和识别率不高的问题，本文提出了
文章引用：刘希东，孟岩，李国友，唐磊，[J]・仪器与设2020, 8(1): 8-20. DOI: