稀疏保持典型相关分析方法研究
时间:2018-03-07 10:54 来源:毕业论文 作者:毕业论文 点击:次
| 摘要在模式识别中，特征抽取与融合一直是研究的热门领域。典型相关分析是相关投影分析的基本方法之一，现已成功应用于各个领域。同时，典型相关分析也被应用到多特征抽取与融合中，并在图像分类识别中获得了良好的效果。 随着压缩感知理论的提出，稀疏表示思想得到越来越的关注并被引入到模式识别领域。稀疏保持典型相关分析由此提出，其核心思想是将稀疏保持项作为一种自然鉴别信息引入到典型相关分析框架中，在实现两组特征有效融合的同时保证特征间稀疏重构约束，进一步加强特征的表示和鉴别能力。本文主要实现是典型相关分析和稀疏保持典型相关分析方法并在相应的数据集上进行实验，验证了稀疏保持典型相关分析方法的优越性和稳定性。19356 关键字 典型相关分析，稀疏保持，稀疏保持典型相关分析，特征融合，文度约减，图像识别 毕业论文设计说明书（论文）外文摘要 Title Research of Sparsity Preserving Canonical Correlation Analysis Abstract Feature extraction and fusion are hot research areas in pattern recognition. Canonical correlation analysis (CCA) is one of the most popular methods in correlation analysis field. CCA has been widely and successfully applied in many fields including multiple feature extraction and fusion. Good recognition results have been achieved in the application of image classification. With rapid development of compressed sensing theory, sparse representation receives more and more attention and has been successfully applied in pattern recognition. Inspired with the success of compressed sensing, sparsity preserving canonical correlation analysis (SPCCA) has been proposed. In SPCCA, the sparsity preserving item, as a kind of nature discriminating information, is incorporated into CCA framework. SPCCA not only fuses the discriminative information of two feature sets but also constrains the sparse reconstructive relationship among each feature set. Therefore, compared with CCA, SPCCA enhances the power of feature representation and obtains better discrimination capability of extracted features. Keywords canonical correlation analysis, sparsity preserving, sparsity preserving, canonical correlation analysis, dimensionality reduction feature fusion, image recognition 目 录 1 绪论 1 1.1 引言 1 1.2 典型相关分析研究现状 3 1.3 稀疏表示研究现状 4 1.4 基于投影分析的手写体识别 4 1.5 本章小结 5 2 典型相关分析及其在特征融合中的应用 6 2.1 典型相关分析基本理论及模型 6 2.2 典型相关分析在特征融合中的应用 7 2.3 本章小结 8 3 稀疏保持典型相关分析 9 3.1 稀疏保持投影方法 9 3.2 稀疏保持典型相关分析 10 3.3 本章小结 12 4 实验与分析 13 4.1 实验平台 13 4.2 数据集 14 4.3 实验步骤 16 4.4 实验结果分析 18 4.5 本章小结 25 总 结 26 致 谢 27 参 考 文 献 28 附 录 31 1 绪论 1.1 引言 (责任编辑：qin) |