摘要
针对当前传统手势识别技术受环境和手部自身条件干扰较大,如当手腕处存在大袖口或其他干扰物的情况下,难以准确识别手势的问题,提出一种基于深度数据的手势识别方法。首先,通过预处理提取手形;其次,利用提出的N-Iterate、C-Loop判定等方法识别手掌最大内切圆;然后,计算手形所有轮廓点到掌心距离的直方图及其波峰索引,并结合角度提取指尖个数;将得到的3类特征作为改进SVM的输入,映射到高维空间,进行手势0~5的识别。实验结果表明,该方法在复杂背景和手部自干扰等影响下具有较高的识别准确率和实时性,平均准确率提高至98.57%,识别耗时降低至37.923 ms,较大程度提高了识别效率。
Aiming at the problem that the current traditional gesture recognition technology is greatly affected by the the environment and its own hand conditions,it is difficult to extract accurate gesture features when there are cuffs or other interference objects at the wrist.An improved gesture method based based on the depth data is proposed.Firstly,the hand shape is extracted by preprocessing.Secondly,the largest inscribed circle of the palm is identified by the proposed N-Iterate,C-Loop judgment method.Then,the histogram of the distance from all the contour points of the hand shape to the palm and its peak index are calculated,and the number of fingertips combined withthe angle between the fingertip and the adjacent point is extracted.Finally,the obtained 4 types of gesture features are used as the input of the improved SVM,mapped to the high-dimensional space,and the gestures 0-5 are recognized.The experimental results show that the method can achieve high recognition accuracy,real-time performance under complex background and multiple influences,which include interference,the average accuracy has increased to 98.57%,and the time-consuming has decreased to 37.923 ms,which greatly improves the efficiency of recognition.
作者
吴文
李春晓
金宏晖
杨梅
王清悦
WU Wen;LI Chunxiao;JIN Honghui;YANG Mei;WANG Qingyue(College of Information Engineering,Yangzhou University,Yangzhou 225127,China)
出处
《激光杂志》
CAS
北大核心
2023年第3期111-117,共7页
Laser Journal
基金
国家自然科学基金青年项目(No.61901408)
扬州大学科创基金(No.X20220378)。
关键词
手势识别
深度数据
最大内切圆
直方图
指尖个数
SVM
gesture recognition
depth data
the maximum inscribed circle
histogram
the number of fingertips
SVM
