| 动态时间规整算法优化 |
| |
| 引用本文： | 叶科淮,陈 志,王仁杰,史佳成,胡 宸.动态时间规整算法优化[J].教育技术导刊,2021,20(1):132-135. |
| |
| 作者姓名： | 叶科淮 陈 志 王仁杰 史佳成 胡 宸 |
| |
| 作者单位： | 1. 南京邮电大学 计算机学院、软件学院、网络空间安全学院，江苏 南京 210023；2. 南京邮电大学 通信与信息工程学院，江苏 南京 210003 |
| |
| 基金项目： | 江苏省重点研发计划（社会发展）项目（BE2016778，BE2019739）；南京邮电大学科研项目（NY217054）；江苏省大学生创新创业训练计划项目（201910293019Z，SZDG2019019） |
| |
| 摘 要： | 为了解决动态时间规整算法在时间序列长度较长、两段时间序列长度相当时计算效率较低等问题，对动态时间规整增加约束条件，并从压缩时间序列、优化全局约束及修改约束条件等方面进行改进。通过实验，将算法应用于较长的时间序列中。实验结果表明，两段时间序列长度越接近，动态时间规整的时间复杂度越趋于线性，在完全相等时，时间复杂度从传统算法的O（nm）优化至O（n），优化效率最高可达到99%。修改约束条件后的动态规整算法可识别实验中所有经过慢放处理的时间序列。经过优化的动态时间规整算法可在一定条件下提高效率，并且能快速识别相同的时间序列。
|
| 关 键 词： | 动态时间规整 动态规划 语音识别 手势识别 数据挖掘 |
| 收稿时间： | 2020-06-02 |
Optimization of Dynamic Time Warping Algorithm |
| |
| YE Ke-huai,CHEN Zhi,WANG Ren-jie,SHI Jia-cheng,HU Chen.Optimization of Dynamic Time Warping Algorithm[J].Introduction of Educational Technology,2021,20(1):132-135. |
| |
| Authors: | YE Ke-huai CHEN Zhi WANG Ren-jie SHI Jia-cheng HU Chen |
| |
| Institution: | 1. School of Computer Science， Nanjing University of Posts and Telecommunications， Nanjing 210023， China；2. College of Telecommunications and Information Engineering， Nanjing University of Posts and Telecommunications， Nanjing 210003， China |
| |
| Abstract: | In order to solve the problem of dynamic time warping algorithm where the length of the time series is long and the length of the two time series is equivalent，causing the low calculation efficiency， restrictions are added to dynamic time warping， and improvements are made in terms of compressing time series， optimizing global constraints， and modifying constraints. Through experiments， the algorithm is applied to longer time series. The experimental results show that the closer the lengths of the two time series are， the more linear the time complexity of dynamic time warping is. When they are completely equal， the time complexity of the traditional algorithm is optimized from O（nm） to O（n）. The optimization efficiency can reach up to 99%. After modifying the dynamic regularization algorithm， the constraints can identify all the time series in the experiment that have been processed slowly. The optimized dynamic time warping algorithm can improve efficiency under certain conditions， and can quickly identify the same time series. |
| |
| Keywords: | dynamic time warping dynamic programming voice recognition gesture recognition data mining |
|
| 点击此处可从《教育技术导刊》浏览原始摘要信息 |
| 点击此处可从《教育技术导刊》下载免费的PDF全文 |
