摘要
针对利用单类特征对不规范书写坐姿进行分类识别率偏低的现状,提出多类特征融合的不规范书写坐姿分类方法。参照规范书写坐姿,不规范书写坐姿归纳为趴写、驼背等7类,应用YCbCr平面投影的椭圆特性,提取各类坐姿在不同亮度下的肤色特征,依据不同阈值进行坐姿的SURF特征提取,对同类坐姿进行异类特征加权融合。BP神经网络分类实验结果表明,该方法的不规范书写坐姿识别率比单类特征法有明显提高,具有更好的实用性。
To solve the problem of low rate when recognizing incorrect writing sitting posture using single feature,the way of u-sing the multi-feature fusion to identify incorrect sitting posture was proposed.According to standard writing posture,incorrect sitting posture for writing induction was classified into seven categories,such as lying to write,humpback and so on.The skin-color features at various brightness were extracted by the application of the clustered skin area in a fixed region of YCbCr space which had ellipse-like projection on CbCr plane.The speeded-up robust features(SURF)of sitting postures were extracted based on different threshold values.The same kind of sitting postures were processed with weighted fusion of different features in similar posture.The multi-class feature,enjoying an higher practical value,is apparently superior in identifying incorrect sit-ting posture,which was showed in the BP neural network classification experiments.
作者
袁迪波
戴永
陈统乾
YUAN Di-bo;DAI Yong;CHEN Tong-qian(Ministry of Education Key Laboratory of Intelligent Computing and Information Processing,Xiangtan University, Xiangtan 411105, China)
出处
《计算机工程与设计》
北大核心
2017年第2期528-533,共6页
Computer Engineering and Design
基金
湖南省教育厅基金项目(13C914)
湖南省十二五重点学科建设基金项目(09K040)
关键词
不规范书写坐姿
多类特征融合
识别
肤色特征
SURF特征
神经网络分类
incorrect writing posture
multi-feature fusion
recognition
skin-color feature
SURF feature
neural network classification
