| EM算法的改进及其在行为识别中的应用 |
| |
| 引用本文： | 赵桂儒,刘典婷. EM算法的改进及其在行为识别中的应用[J]. 电视技术, 2014, 38(13) |
| |
| 作者姓名： | 赵桂儒 刘典婷 |
| |
| 作者单位： | 中国地震台网中心,迈阿密大学 |
| |
| 基金项目： | 国家留学基金资助项目；地震行业专项 |
| |
| 摘 要： | EM算法是求解GMM参数的传统算法,当样本数据规模比较大、GMM高斯成分数量比较高时,EM算法需要很长的时间才能收敛。提出了一种改进的EM算法,通过设置适当的参数,利用改进后的EM算法求解GMM参数,相比原EM算法在运行速度上有了很大的提高;进一步,结合GMM超向量以及SVM分类器,将改进后的EM算法应用到对KTH人体行为数据库的识别中,相比原EM算法识别准确率只受到了很小的影响。
|
| 关 键 词： | 高斯混合分布 EM算法 行为识别 |
| 收稿时间： | 2014-02-17 |
| 修稿时间： | 2014-03-26 |
The Application of Improved EM Algorithm in Recognition of Human Actions |
| |
| ZHAO GUIRU and Liu Dianting. The Application of Improved EM Algorithm in Recognition of Human Actions[J]. Ideo Engineering, 2014, 38(13) |
| |
| Authors: | ZHAO GUIRU and Liu Dianting |
| |
| Affiliation: | China Earthquake Networks Center,University of Miami |
| |
| Abstract: | A long time is needed to estimate GMM parameters using the general EM algorithm when the training data is large and the number of Gauss components is big. An improved EM algorithm is proposed in this paper. Compared to the original algorithm the operating speed is significantly improved using the improved EM algorithm to estimate GMM parameters if the appropriate parameters are given; next, the improved EM algorithm is applied on the KTH human actions database combined with GMM supervector and SVM, the recognition accuracy is affected only a little compared to the original EM algorithm. |
| |
| Keywords: | GMM EM recognition of human action |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《电视技术》浏览原始摘要信息 |
|
点击此处可从《电视技术》下载全文 |
|
