| 基于深度图像HOG特征的实时手势识别方法 |
| |
| 引用本文： | VanBang L E,朱煜,赵江坤,陈宁.基于深度图像HOG特征的实时手势识别方法[J].医学教育探索,2015(5):698-702. |
| |
| 作者姓名： | VanBang L E 朱煜 赵江坤 陈宁 |
| |
| 作者单位： | 华东理工大学信息科学与工程学院, 上海 200237,华东理工大学信息科学与工程学院, 上海 200237,华东理工大学信息科学与工程学院, 上海 200237,华东理工大学信息科学与工程学院, 上海 200237 |
| |
| 基金项目： | 国家自然科学基金(61271349);中央高校基本科研业务费专项资金(WH1214015) |
| |
| 摘 要： | 手势识别是模式识别领域的一个热点研究方向。提出了一种利用Kinect传感器深度图像进行手势分割的方法,并研究了基于灰度图像HOG特征的手势识别模型;深入研究了HOG特征,分析其特征向量特点,探讨了不同特征维数对训练机的影响及处理效率;通过SVM机器学习方法实现手势的分类识别,经过对大量实验样本的优化训练,获得了最优SVM参数,并进行分析、对比识别率。本文方法维数少、识别率高、运行速度快、性能稳定,能满足实时性手势识别的要求。
|
| 关 键 词： | Kinect 深度图像 HOG特征 SVM机器学习 手势识别 |
| 收稿时间： | 2014/12/1 0:00:00 |
Real-Time Gesture Recognition Method Based on Depth Image HOG Features |
| |
| VanBang L E,ZHU Yu,ZHAO Jiang-kun and CHEN Ning.Real-Time Gesture Recognition Method Based on Depth Image HOG Features[J].Researches in Medical Education,2015(5):698-702. |
| |
| Authors: | VanBang L E ZHU Yu ZHAO Jiang-kun and CHEN Ning |
| |
| Institution: | School of Information Science and Engineering, East China University of Science and Technology, Shanghai 200237, China,School of Information Science and Engineering, East China University of Science and Technology, Shanghai 200237, China,School of Information Science and Engineering, East China University of Science and Technology, Shanghai 200237, China and School of Information Science and Engineering, East China University of Science and Technology, Shanghai 200237, China |
| |
| Abstract: | Gesture recognition is a hot topic in the field of pattern recognition. By means of depth information of Kinect sensor, this paper proposes a gesture segmentation method and a gesture recognition model based on HOG features of grayscale image. Besides, this paper researches HOG features, analyzes the characteristics of the eigenvectors, and explores the influence of different feature dimensions on training machine and processing efficiency. Finally, SVM method is utilized to realize the classification of gesture recognition, in which the SVM parameters are optimized through a large number of experiment data and the rate of identification is compared and analyzed. It is shown that the proposed method has less dimension, high recognition rate, quick running, and stable performance such that it can meet the requirements of real-time gesture recognition. |
| |
| Keywords: | Kinect depth image HOG features SVM machine learning gesture recognition |
|
| 点击此处可从《医学教育探索》浏览原始摘要信息 |
| 点击此处可从《医学教育探索》下载免费的PDF全文 |
|
