摘要
针对高精度的实时人体姿态识别,选取简化特征提取运算的时域特征作为唯一特征量;运用基于模糊模式识别的算法,建立一种高效的跨平台的C/S网络架构,对人体姿态进行实时识别;构造支持向量机多类分类器,运用独立检测法对识别结果进行精度评价,建立混淆矩阵对姿态之间的相似性进行度量。实验结果表明,该系统适用于普通资源受限的智能手机,具有较好的通用性,对于人体姿态的实时识别具有较好的效果,消耗内存资源较少,具有良好的实用价值。
To achieve the real-time human activity pattern recognition with high accuracy,the simplify calculation of the feature extraction of domain features was selected as the only features.Using fuzzy pattern recognition method,an efficient C/S network based on different platforms was approached to identify real-time human postures.SVM was established,independent test method was evaluated and similarity measurement was used to depend on confusion-matrix at last.The results show that this algorithm can effectively run on resource-constrained ordinary smartphone with good versatility.At the same time,for the realtime human activity pattern recognition,it can achieve better effectiveness and also consume less memory resources,it is suitable for practical applications.
作者
王玉坤
高炜欣
王征
武晓朦
WANG Yu-kun GAO Wei-xin WANG Zheng WU Xiao-meng(College of Electronic Engineering, Xi＇an Shiyou University, Xi＇an 710065, China)
出处
《计算机工程与设计》
北大核心
2016年第11期3092-3096,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(E040901)
2013陕西省自然科学基金项目(2013JQ8049)
陕西省自然科学基础研究计划青年人才基金项目(2015JQ5129)
中国石油科技创新基金研究项目(2014D-5006-0605)
关键词
跨平台
时域
模糊识别
混淆矩阵
智能手机
different platforms
time domain
fuzzy pattern
confusion-matrix
smartphone
