摘要
词袋模型的复杂度高,且区分能力较弱,为解决这一问题,在经典词袋模型的基础上,提出一种有监督的词袋模型。在训练过程中对训练样本类别进行标记,在此基础上构建直方图总体能量目标函数,依据能量最小准则学习码本。通过文本检索和图像检索两组多媒体信息检索实验进行对比,对比结果表明,有监督词袋模型比经典词袋模型的检索精确度高、检索耗时少。
The training and coding process of the classic bag of word model is unsupervised and does not require tag data.Although the adaptability of this approach is strong,the bag of word model is highly complex and has a weak ability to distinguish.To solve this problem,a supervised bag of word model based on the classical one was put forward.The proposed model needed to mark the category of samples in the training process.On this basis,the objective function of the overall energy of histogram was constructed,and the codebook was learned according to the minimum energy criterion.Through experimental comparison on two groups of multimedia information retrieval experiments including text retrieval and image retrieval,the results show that the supervised bag of word model is more accurate and less time-consuming than the classical one.
作者
袁桂霞
周先春
YUAN Gui-xia;ZHOU Xian-chun(School of Information and Mechanical and Electrical Engineering,Jiangsu Open University,Nanjing 210017,China;School of Electronic and Information Engineering,Nanjing University of Information Science and Technology,Nanjing 210017,China)
出处
《计算机工程与设计》
北大核心
2018年第9期2873-2878,共6页
Computer Engineering and Design
基金
国家创新基金项目(435012C26244104350)
关键词
词袋模型
多媒体信息检索
文本检索
图像检索
能量最小准则
bag of word model
multimedia information retrieval
text retrieval
image retrieval
minimum energy criterion
