Myo手势识别方法在古建筑漫游系统中的应用  

Application of Myo Gesture Recognition Method in Ancient Building Roaming System

作　　者：薛艳萍 王学松[1] 武仲科 王醒策 周明全 Xue Yanping;Wang Xuesong;Wu Zhongke;Wang Xingce;Zhou Mingquan(Institute of Virtual Reality & Visual Technology,College of Information Science and Technology,Beijing Normal University,Beijing 100875,China)

机构地区：[1]北京师范大学信息科学与技术学院虚拟现实与可视化技术研究所

出　　处：《系统仿真学报》2019年第9期1907-1915,共9页Journal of System Simulation

基　　金：国家重点研发计划(2017YFE0100500);国家科技支撑计划(2017YFB1402105,2017YFB1002604,2017YFB1002804);北京市自然科学基金(4172033)

摘　　要：博物馆展品众多,但展示方式单一,多用文字图片展现古建筑,无法给游客逼真体验,由此提出基于Myo臂环结合虚拟现实的古建筑漫游,以开封铁塔公园为例实现整体系统。设计得到游客佩戴Myo完成前进、转向、下蹲、停止动作时的加速度和陀螺仪值,设计SVM手势识别算法对特征值分类,设计对比实验验证算法的识别准确率和效率。建立特征值数据库,绘制其在固定时间内变化折线图,并用T-分布法求出置信区间。网格搜索和K-折验证法验证识别准确性,在动态环境中可达95%的准确率。系统实现对用户手势识别,提升用户与场景的交互性,可广泛应用到类似系统。Museum display many exhibits, but its display way is single.The ancient buildings are displayed by texts and pictures, so it cannot bring a realistic experience to the visitors.Therefore, the ancient architectural scene roaming based on Myo combined with virtual reality is proposed.Kaifeng Tower Park is used as an example to achieve the system.In the design, system gets visitors’ acceleration and gyroscope values of forward, turn, squat and stop by wearing Myo.SVM-based gesture recognition algorithm is designed to complete the classification of feature values.Contrast experiment is designed to verify accuracy and efficiency of algorithm.Database is established to draw a line graph of its change in a fixed time.T-distribution method is used to obtain the confidence interval.Grid search and K-fold verification method are used to test the gesture recognition.95% accuracy can be achieved in a dynamic environment.The system realizes the recognition of user gestures and improves users’ interaction with the scene.It can be widely applied to similar systems.

关 键 词：古建筑模拟 手势识别 虚拟漫游 人机交互 MYO 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]