无线视频传感器网络的分布式采样和适配化传输

项目名称： 无线视频传感器网络的分布式采样和适配化传输

项目编号： No.61271211

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 无线电电子学、电信技术

项目作者： 邹君妮

作者单位： 上海大学

项目金额： 75万元

中文摘要： 无线视频传感器网络，受限于资源约束和视频信息的高维冗余相关，结合网络和信号拓扑的稀疏采样和高效传输，成为时下重要的理论与应用课题。本申请拟基于统计学习和泛函优化的方法，建立自适应的字典和观测矩阵，利用高维视频信息的稀疏性与相关性，去除视间-时间-空间的高阶冗余，优化数据表达。同时，研究视频信息协作采样与通信的稀疏投影模型，进行无线中继协作通信的均衡功率分配，实现协作通信网络信息流空间上的信源-网络联合鲁棒优化，逼近传感网络的理论性能，建立相应的视频信息压缩采样与传输系统平台。提出基于分布式视频源联合稀疏表示模型的优化学习方法，把传统低维理论推广到多维度领域；将稀疏投影过程与网络拓扑结构、动态路由关联起来，协同地完成对信号的压缩与汇聚，建立投影稀疏度、观测次数、重构误差三者之间关系模型。预期本项目研究成果，能应用于高性能无线视频传感网，在掌握自主知识产权的同时,我国在这一研究领域居国际前沿。

中文关键词： 无线视频传感器网络；分布式视频；稀疏采样；协作传输；资源分配

英文摘要： Due to resource constraints and information redundancy, sparse sampling and efficient transmission of video streaming on the basis of netowrk and signal structure emerges a promising topic in wireless video sensor networks. We attempt to apply statistic learnig and functional analysis method in constructing adaptive dictionary and measurement matrix, remove high dimensional redundancy by exploiting the sparsity and corelation of video data. Meanwhile, we study the sparse projection model for cooperatively sampling and transmission of video streaming, address auction-based power allocation in wirless cooperative communications, and perform a joint source and network robust optimization for maximizing the overall throughput of the wireless sensor network. Also, we try to establish a testbed for realizing sparse sampling and transmission of video information. Our main contributions include: A optimized learning method to joint sparse represenation modeling for distributed video sources, which extends conventional theory for low dimentional signals to that for high dimentional signals; Combining sparse projection with network topology and dynamic routing, which cooperatively fulfills signal compression and convergence; Modeling the relationship of the sparsity degree, the number of measurements, and the reconstructi

英文关键词： wireless video sensor networks；distributed video；sparse sampling；cooperative communication；resource allocation