查询结果:   李云峰,张澎悦.基于GLCM和Gabor纹理特征的手势识别算法[J].计算机应用与软件,2019,36(7):183 - 191.
中文标题
基于GLCM和Gabor纹理特征的手势识别算法
发表栏目
人工智能与识别
摘要点击数
670
英文标题
GESTURE RECOGNITION ALGORITHM BASED ON TEXTURE FEATURE OF GLCM AND GABOR
作 者
李云峰 张澎悦 Li Yunfeng Zhang Pengyue
作者单位
河南科技大学机电工程学院 河南 洛阳 471003 机械装备先进制造河南省协同创新中心 河南 洛阳 471003    
英文单位
School of Mechatronics Engineering, Henan University of Science and Technology, Luoyang 471003,Henan,China Collaborative Innovation Center of Machinery Equipment Advanced Manufacturing of Henan Province , Luoyang 471003,Henan,China    
关键词
手势识别 纹理特征 共生矩阵 Gabor小波变换 深度堆栈自编码网络
Keywords
Gesture recognition Texture feature Co-occurrence matrix Gabor wavelet transform Deep stack encoder network
基金项目
国家自然科学基金项目(61702163)
作者资料
李云峰,副教授,主研领域:生物特征识别,机器学习。张澎悦,硕士生。 。
文章摘要
针对手势灰度图像的纹理特征富含手势类别信息的特点,提出一种基于融合GLCM(灰度共生矩阵)和Gabor小波变换提取手势图像空、频域纹理特征的手势识别方法。构建手势灰度图像的多方向共生矩阵,并计算多方向共生矩阵的特征参数来提取手势纹理的GLCM特征;通过手势灰度图像的Gabor小波变换来提取手势纹理的Gabor特征;对所提取的两种特征进行归一化处理后串联构建手势纹理特征向量;使用基于稀疏自动编码器和softmax分类器的深度堆栈自编码网络对构建的手势纹理特征向量进行分类识别。实验表明:该方法具有较高的识别率和较好的鲁棒性,对15种手势的平均识别率达到97.4%,能够满足人机交互对手势识别的要求。
Abstract
For the characteristic that the texture feature of gray gesture image is rich in the information of gesture category, a gesture recognition method based on fusion of Gray-level Co-occurrence Matrix(GLCM) and Gabor wavelet transform was proposed to extract spatial and frequency domain texture features of gesture images. We, constructed multi-direction co-occurrence matrix of gesture gray image and calculated feature parameters of multi-direction co-occurrence matrix to extract GLCM features of gesture texture. The Gabor features of gesture texture was extracted by Gabor wavelet transform of gray gesture image. Then, the two extracted features were normalized to construct the feature vector of gesture texture in series. The deep stack encoder network based on sparse encoder and softmax classifier was used to classify and recognize feature vector of gesture texture. Experiments show that the proposed method has high recognition rate and good robustness, with the average recognition rate of 97.4% among fifteen gestures, which can meet the requirements of gesture recognition for human-computer interaction.
下载PDF全文