模拟电路干扰信号盲分离的独立成分分析方法  

Independent component analysis method for blind separating interference signals in analog circuits

作　　者：李昊[1] 董永贵[1] 

机构地区：[1]清华大学精密仪器系精密测试技术及仪器国家重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2015年第3期339-344,共6页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(61271129)

摘　　要：为在模拟电路调试阶段分析干扰信号的来源,构建了一种基于瞬态线性混合模型的分析方法。将干扰信号视为统计独立的源信号,由多个电路节点的测试信号组成观测向量,利用独立成分分析算法实现了干扰信号的盲分离。考虑到测试电路中的输入激励为已知确定性信号,提出了一种在观测向量中增加已知输入信号以改善分离效果的方法。以信号失真比作为评价指标,研究了电容等储能电路元件对分离效果的影响。数值仿真计算与实际电路测试结果表明:储能元件在电路中的卷积效应,是影响分离效果的重要因素。在观测向量中添加已知输入信号,可显著改善干扰信号的分离效果。An instantaneous linear mixing model was developed to analyze the sources of interference signals during analog circuit adjustment.The interference signals were modeled as statistically independent sources.The observed vectors were constructed from signals measured at several circuit nodes.The blind separation of the interference signals used an independent component analysis algorithm.Since the input excitation to the test circuit was a known deterministic signal,the known signal input signal can be included in the observed vectors to improve the separation performance.The separation performance was studied using the source to distortion ratio as an indicator for the influence of the energy storage circuit elements,such as the capacitances,on the signal separation.Numerical simulations and experimental measurements show that the convolution effect of the energy storage elements in the circuits is an important factor that reduces the separation efficiency.If the known input signals are included in the observed vectors,the interference signal separation can be significantly improved.

关 键 词：盲源分离 独立成分分析 电磁兼容 干扰信号 模拟电路 

分 类 号：TN911.7[电子电信—通信与信息系统]