项目名称: 稀疏表示结合质量评价的多模生物特征识别研究
项目编号: No.61305008
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 自动化技术、计算机技术
项目作者: 王风华
作者单位: 中国石油大学(华东)
项目金额: 23万元
中文摘要: 作为未来信息安全的重要解决方案,多模生物识别技术的研究受到广泛关注。本项目拟借助稀疏表示理论,以人脸、虹膜和人耳作为实验对象,从特征级融合入手,对多模生物识别展开新方法研究。项目的核心思路是将人类感知图像的稀疏性机制与多模生物识别研究结合起来,把多模数据的特征级融合识别转化为稀疏模型的求解。项目的预期成果包括:(1)通过研究多模稀疏表示模型的构建、有监督的字典学习及多模稀疏模型求解等内容,提出多模稀疏表示的特征级融合识别算法。(2)针对多模生物识别中的非线性可分问题,借助核技巧,引入非线性表示,提出基于核稀疏表示的多模生物特征识别算法。(3)针对环境复杂导致的采集样本的质量变化问题,引入质量评价机制,构建质量评价约束的多模核稀疏表示模型,提出核稀疏表示结合质量评价的多模生物识别理论框架及相应算法。本项目的开展将为多模生物识别的研究提供新的思路和方法,有助于推动多模生物识别产品实用化进展。
中文关键词: 多模生物特征识别;稀疏表示;质量评价;数字水印;
英文摘要: Research on multimodal biometrics,which is an important solution in future information security technology,receives extensive attentions.This proposal will study the new multimodal biometric methods via sparse representation theory at the feature level, integrating face, iris and ear. The core thinking is to transform fusion of multimodal data at the feature level to the solving of sparse representation models by linking sparsity of human vision perception and multimodal biometrics.The expected achievements of this proposal include: (1)A multimodal biometric algorithm based on multimodal sparse representation will be proposed by studying the building of multimodal sparse representation model, supervised dictionary learning methods and the solving of sparse model.(2) Aiming at the non-linear classification problems, kernel trick is used, and a multimodal biometric authentication algorithm based on kernel sparse representation model will be proposed.(3) The diversity of surroundings and the instability of devices make the quality of sample image uncertain. To solve this problem, a multimodal biometric authentication algorithm based on kernel sparse representation and quality assessment will be proposed, which has better self-adaptive ability for surroundings. The research of proposed project provides a new thread
英文关键词: multimodal biometrics;sparse representation;quality assessment;digital watermarking;