摘要
手势识别是人机交互领域的一种重要手段,针对手势形态多样性和背景的复杂性导致识别率不高的问题,提出一种融合HOG+SVM的手势识别方法,该方法有效提高手势识别率。首先建立手势样本数据集,选择轮廓信息完整的手势图像作为模板,为了验证分割的高效性,采集6类手势的6,000张样本,提取两种局部二值模式特征和一种方向梯度直方图,对形态学处理后手势样本集提取HOG特征并进行降维处理,目的是提高手势识别速度,然后对手势轮廓和质心位置提取不同形态手势多特征信息,对两种特征进行归一化处理,精确地对手势信息进行识别,得到不同形态手势的特征,将最终的手势分类特征通过SVM进行分类识别。实验结果表明,本文提出的手势识别方法在复杂环境下识别率达到95%,具有较强的鲁棒性,满足人机交互的需求。
Gesture recognition is an important method in the field of human-computer interaction.Aiming at the problem of low recognition rate caused by the diversity of gesture morphology and the complexity of the background,the paper proposes a gesture recognition method combining HOG+SVM.First,establish a gesture sample data set,extract HOG features from the morphologically processed gesture sample set and perform dimensionality reduction processing,then extract multi-feature information of different morphological gestures from the gesture contour and centroid position,and normalize the two features.Obtain the features of different gestures,and classify and recognize the final gesture classification features through SVM.The experimental results show that the gesture recognition method proposed in the paper has a recognition rate of 95%in a complex environment,and has strong robustness to meet the needs of human-computer interaction.
作者
李国玄
马凯凯
王文博
LI Guoxuan;MA Kaikai;WANG Wenbo(School of Mechanical Engineering,Shangqiu Insititute of Technology,Shangqiu 476000,China)
出处
《传感器世界》
2022年第12期30-36,I0003,共8页
Sensor World
基金
商丘工学院科研项目(No.2022KYXM22)。
关键词
手势识别
HOG
SVM
多特征信息
人机交互
gesture recognition
HOG
SVM
multi-feature information
human-computer interaction
