摘要
常用的手势识别方法受限于有限的二维图像信息,难以从复杂的背景中有效地分割出目标像素,且大多依赖于监督学习的分类方法,只能在有限的手势库中进行选择判定,无法适用于较为精细的手指运动检测。为此,提出一种利用粒子群寻优算法来估计手指关节角度的方法,引入Kinect深度图像优化特征提取,提高检测准确性,通过对手部自由度的分析,引入多个约束条件,减少需要预测和估计的自由度个数。基于粒子群寻优算法,得出最优的预测模型,将手势分类问题转化为手指关节角度变量求解问题。实验结果表明,该方法有效地提高了手势检测中的检测准确率,降低了检测失效的情况。
Traditional methods of gesture recognition are incompetent to detect finger＇s delicate movement due to poor segmentation effect using 2D images and limited gesture templates through supervised training of classifier.This paper proposes a method of finger joint angle measurement by Particle Swarm Optimization（PSO）algorithm,introduces Kinect depth image to optimize feature extraction and improve accuracy.Through the analysis of hand free degree,it introduces multiple constraints to reduce the degrees of freedom number,optimize PSO to calculate the best model and analyze the measurements,transform the common problems of gesture classification to the variable solution of finger joint angle.Experimental results show that this method can effectively improve the detection accuracy and reduce the situation of detection fault.
出处
《计算机工程》
CAS
CSCD
北大核心
2015年第10期221-225,231,共6页
Computer Engineering
基金
浙江省国际科技合作基金资助项目(2012C34G2020027)
关键词
手势识别
深度图像
肤色检测
粒子群优化算法
手指关节角度
gesture recognition
depth image
skin color detection
Particle Swarm Optimization（PSO）algorithm
finger joint angle
