基于支持向量机的人机交互媒体播放界面手势识别方法  

Gesture Recognition Method of Human-computer Interaction Media Playing Interface Based on Support Vector Machine

作　　者：张琳钦[1] ZHANG Linqin(School of Computer and Art,Anhui Industrial Economics Vocational and Technical College,Hefei 230051,China)

机构地区：[1]安徽工业经济职业技术学院计算机与艺术学院,合肥230051

出　　处：《西安航空学院学报》2023年第3期83-88,共6页Journal of Xi’an Aeronautical Institute

基　　金：安徽省双基示范双基课程项目(2020SJJXSFK0446);安徽省高校人文社科研究一般项目(2021sk13)。

摘　　要：为了提高手势识别在人机交互媒体播放界面交流的准确率,提出一种基于支持向量机的人机交互媒体播放界面手势识别方法。首先通过分析人机交互媒体播放界面,利用Cam Shift方法跟踪用户手势,确定手势在人机交互媒体播放界面的区域,通过计算波峰数量、手势的长度和能量,提取出手势特征;然后通过扫描用户手势的深度图像,得到手势范围,引入高斯滤波函数,过滤掉手势图像的掩模,确定手心位置,完成手势分割;最后利用支持向量机的分类阈值,计算手势图像的分类面,引入拉格朗日算法,将最优分类面问题转化为对偶性问题,实现人机交互媒体播放界面的手势识别。实验结果表明,该方法能够清晰识别人机交互媒体播放界面中的手势,并将准确率提高到90%以上。研究结果为复杂工作环境中运用手势动作进行人机互动提供理论支持。In order to improve the communication accuracy of gesture recognition in human-computer interactive media playback interface,a gesture recognition method based on support vector machine is proposed.Firstly,by analyzing the interface of human-computer interactive media playback,the user gestures are tracked by Cam Shift method to determine the area of gestures in the interface of human-computer interactive media playback,and the characteristics of gestures are extracted by calculating the number of peaks,length and energy of gestures.Then,by scanning the depth image of users′gestures,the range of gestures is obtained.The Gaussian filter function is introduced to filter out the mask of gesture images and determine the palm position to complete the gesture segmentation.Finally,the classification surface of gesture images is calculated by using the classification threshold of support vector machine,and the Lagrangian algorithm is introduced to transform the problem of optimal classification surface into a duality problem,so as to realize the gesture recognition of interactive media playback interface.Experimental results show that this method can clearly identify gestures in the interface of human-computer interactive media playback,and the accuracy rate is higher than 90%.The results of this study provide theoretical support for human-computer interaction with gestures in complex work environments.

关 键 词：支持向量机 人机交互 手势识别 手势特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]