利用参量结构实现复杂信号环境下盲信号分离方法研究

项目名称： 利用参量结构实现复杂信号环境下盲信号分离方法研究

项目编号： No.61201407

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 电子学与信息系统

项目作者： 徐先峰

作者单位： 长安大学

项目金额： 26万元

中文摘要： 针对复杂信号环境下目标信号的提取问题，着眼于机械系统状态监测及故障诊断等实际应用领域，采用盲信号分离技术作为解决方法，在前期研究的基础上，对密集信号进行稀疏化预处理，并构造高精度降维矩阵以达到数据降维及噪声抑制，深入研究复杂信号环境的典型特点建立科学合理信号模型，充分挖掘利用观测数据的空域、时域、频域信息以及观测数据所有可能的参量结构，设计有效代价函数，在实现非凸优化问题松弛的基础上，利用先进的凸优化算法，研究高效盲信号分离方法。在优化过程中，巧妙利用参量结构的先验信息，并及时发掘利用中间参量新的结构特点，采用多阶段分解与重构的思想，研究计算复杂度低、收敛速度快、稳健性能好且适合欠定、小样本、低信噪比等复杂情况的快速迭代算法，实现复杂信号环境下目标信号的盲分离。本项目的研究成果将为盲信号分离技术的深入发展及其在机械系统等复杂信号环境领域的实际应用提供坚实的理论基础和强大的技术支持。

中文关键词： 盲信号分离；波达方向估计；机械故障诊断；目标跟踪；无线传感器网络

英文摘要： The blind source separation (BSS) technique is proposed to retrieve the target source signals from mixtures captured in complicated signal environment, for the purpose of its application in such realistic field as machine monitoring and damage detection in mechanical system. Based on the existed research foundations, the dense mixtures will be made sparse. High quality matrix will be developed to low the data dimension and restrain the noise. The traits of complicated signal environment will be deeply studied in order to establish the logical signal model. The cost function will be designed on the base of the full utilization of the structural characters of concerned variables and the space-domain, time-domain and frequency-domain information comprised by received mixtures. After the relaxation procedure of non-convex problem, the convex optimization algorithm is utilized to derive the BSS algorithm. During the process of optimization, the ingenious utilization of prior information and the new structural characters, and the adoption of the method of multi-stage decomposition and integration, will result with the good iterative algorithm which posses the traits like low computation, fast convergence, good robustness and fitting to complicated signal environment such as underdetermined, a few snapshots and low sig

英文关键词： Blind Signal Separation (BSS)；Direction-of-Arrival (DOA)；Machine Damage Detection；Target Tracking；Wireless Sensor Networks (WSN)