项目名称: 基于深度学习的海量截获卫星数据分析技术研究
项目编号: No.61501306
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 无线电电子学、电信技术
项目作者: 关庆阳
作者单位: 沈阳航空航天大学
项目金额: 16万元
中文摘要: 为了克服海量截获卫星数据分析面临的技术瓶颈,将深度学习引入海量截获卫星数据分析中。通过研究基于DBN网络的深度学习架构,基于超完备基的稀疏编码方式,以及基于模糊函数的深度学习决策,构建海量数据潜在的分布多层复杂特征表达,用海量数据去学习特征,先从数据去学习较低层次特征,然后在这些低层次特征的基础上再建立更高层次表达,最终形成海量数据的线性或者非线性组合特征表达,以超越人工提取数据特征表达的极限,不仅能获取卫星传输方式、调制方式、编码方式等较低层次的情报信息,也能获取卫星的工作模式、行为特征、协作方式、战术意图等较高层次情报信息,从而提高海量截获卫星数据分析精度,挖掘隐藏海量数据的情报信息,并使所研究的成果在将来能够运用于海量截获卫星数据分析中。
中文关键词: 卫星通信;数据分析;深度学习;稀疏编码;模糊函数
英文摘要: In order to overcome technical bottlenecks of massive interceptive satellite data analysis, deep learning has been introduced into the analysis for massive interceptive data. Through the design for deep learning architecture based on DBN network, sparse coding method based on perfect basis vector, and deep learning decision based on fuzzy function, to construct potential distributed multiple complex expression algorithm from massive data, to learn the low level features from the mass of the satellite data, and to build these low level features from mass data, then to build higher expression based on these low level features, to repeat these low levels of expression form linear or nonlinear combinations of expression, in order to surpass artificial feature extraction limitation, not only to get the way of communication, modulation mode, behavior characteristics, coding method of lower level information, but also to get the way of communication, working mode, behavior characteristics, the cooperation way, tactical intention of higher level information so as to improve the accuracy for interceptive satellite data analysis, to mine the hidden mass data information, and the research results can be applied for massive interceptive satellite data analysis.
英文关键词: satellite communication;data analysis;deep learning ;sparse coding ;fuzzy function