Current position: Home >> Scientific Research >> Paper Publications

基于Self-Training和HMM的时间序列分类

Release Time:2019-03-09  Hits:

Indexed by: Conference Paper

Date of Publication: 2010-09-01

Page Number: 56-59,71

Key Words: 自训练;隐马尔可夫模型;半监督;动态弯曲路径

Abstract: 提出了一种基于Self-Training和隐马尔可夫模型的时间序列分类方法,采用基于动态弯曲路径距离的Self-Training最近邻方法采扩大标记时间序列数据集,利用扩大的标记数据初始化HMM,并且使用基于半监督的迭代学习来进一步训练HMM。实验结果表明该方法能有效避免少量标记数据下HMM参数估计不准确的问题,提高了分类的准确率。

Prev One:A novel access control strategy for distributed data systems

Next One:Face recognition based on local binary patterns with threshold