Consider a real LTI system with a WSS process $x(t)$ as input and WSS process $y(t)$ as output. Base on the WSS correlation properties,we get these equations

$\begin{align*}
&Time-Domain  &:&R_{yy}(\tau) &= h(\tau)*h(-\tau)*R_{xx}(\tau)\\
&Frequency-Domain &:&S_{yy}(j\omega) &= H(j\omega)H^*(j\omega)S_{xx}(j\omega)
\end{align*}$

The way we get $x(t)$ from white noise is no different. Let the input be a white noise with PSD $W_{xx}(j\omega)=1$,which means that its auto-correlation is $\delta$. Then the system can be seen to be a modeling filter denoted by $m(t)$ in time-domain and $M_{xx}(j\omega)$ in frequency-domain.

This can be summarized as the following equations

$\begin{align*}
&Time-Domain  &:&R_{xx}(\tau) &= m_{xx}(\tau)*m_{xx}(-\tau)\\
&Frequency-Domain &:&S_{xx}(j\omega) &= M_{xx}(j\omega)M_{xx}^*(j\omega)
\end{align*}$

Now, to think of a system which is the cascade of the filter $m_{xx}(\tau)$ and $m_{xx}(-\tau)$.

The filter $m_{xx}(\tau)$ can be decomposed into the sum of an even part $m_e(\tau)$, and an odd part $m_o(\tau)$

$m_{xx}(\tau) = m_e(\tau)+m_o(\tau)$

where

$\begin{align*}
m_e(\tau)&= \frac{1}{2}(m_{xx}(\tau)+m_{xx}(-\tau))\\
m_o(\tau)&= \frac{1}{2}(m_{xx}(\tau)-m_{xx}(-\tau))\\
\end{align*}$

If the filter $m_{xx}(\tau)$ is causal, in order that $m_{xx}(\tau)=0$ for $\tau<0$, we require that

$m_o(\tau) = \left\{\begin{matrix}
m_e(\tau), &\tau >0 \\
-m_e(\tau), &\tau<0
\end{matrix}\right.\ =sgn(\tau)m_e(\tau)$

Then the causal impulse response may be written in terms of the even function alone

$\begin{align*}
&m_{xx}(\tau) &= m_e(\tau)+sgn(\tau)m_e(\tau)\\
&m_{xx}(-\tau) &= m_e(\tau)-sgn(\tau)m_e(\tau)
\end{align*}$

For example

In the frequency domain, the frequency response function $M_{xx}(j\omega)$ can also be expressed in terms of the even function alone

$\begin{align*}
M_{xx}(j\omega) &= \mathcal{F}\Big\{m_e(\tau)\Big\}+\mathcal{F}\Big\{sgn(\tau)m_e(\tau)\Big\}\\
&= \mathcal{F}\Big\{m_e(\tau)\Big\}+\frac{1}{2\pi}\mathcal{F}\Big\{sgn(\tau)\Big\}\otimes \mathcal{F}\Big\{m_e(\tau)\Big\}\qquad convolution\ theorem\\
&= M_e(j\omega) + j\left[\frac{1}{\pi\omega}\otimes M_e(j\omega) \right]\\
&= M_e(j\omega) + j\widehat{M}_e(j\omega) \qquad \widehat{M}_e(j\omega)\ means\ Hilbert\ Transform\ of\ M_e(j\omega)
\end{align*}$

The frequency response function $M_{xx}^*(j\omega)$ can be derived with the same argument.

$\displaystyle{M_{xx}^*(j\omega) = M_e(j\omega) - j\widehat{M}_e(j\omega)}$

Thus

$\begin{align*}
S_{xx}(j\omega)&=M_{xx}(j\omega)M_{xx}^*(j\omega)\\
&=\Big\{M_e(j\omega)+j\widehat{M}_e(j\omega)\Big\}\Big\{M_e(j\omega)-j\widehat{M}_e(j\omega)\Big\}\\
&=M_e^2(j\omega)+\widehat{M}_e^2(j\omega)
\end{align*}$

Back to the WSS process, $S_{xx}(j\omega)$ is the PSD of $x(t)$. For real WSS process, the PSD should meet 3 condictions:even, real, non-negative. These condictions can be easily varified on $M_e^2(j\omega)+\widehat{M}_e^2(j\omega)$.

  1. $M_e^2(j\omega)+\widehat{M}_e^2(j\omega)$ is real, because it is the sum of square
  2. $M_e^2(j\omega)+\widehat{M}_e^2(j\omega)$ is non-negative, because it is the sum of square
  3. The first term is the square of FT of real even function, so that $M_e(j\omega)$ is real and even. The second term is the Hilbert transform of the real even function $M_e(j\omega)$. According to the Hilbert transform duality, $\widehat{M}_e(j\omega)$ is odd, which means that $\widehat{M}_e^2(j\omega)$ is even. With these understanding, it is evident that $M_e^2(j\omega)+\widehat{M}_e^2(j\omega)$ is even.

Reference :

MIT Open course 2.161 Signal Processing: Continuous and Discrete: Determining a System's Causality from its Frequency Response

Alan V. Oppenheim: Signals, Systems and Inference, Chapter 11: Wiener Filtering

WSS Process On Causal LTI System的更多相关文章

  1. Create process in UNIX like system

    In UNIX, as we’ve seen, each process is identified by its process identifier, which is a unique inte ...

  2. Linux利器 strace [看出process呼叫哪個system call]

    Linux利器 strace strace常用来跟踪进程执行时的系统调用和所接收的信号. 在Linux世界,进程不能直接访问硬件设备,当进程需要访问硬件设备(比如读取磁盘文件,接收网络数据等等)时,必 ...

  3. Wiener Filter

    假设分别有两个WSS process:$x[n]$,$y[n]$,这两个process之间存在某种关系,并且我们也了解这种关系.现在我们手头上有process $x[n]$,目的是要设计一个LTI系统 ...

  4. LTI系统对WSS Processes的作用

    本文主要专注讨论LTI系统对WSS Process的影响.WSS Process的主要特性有mean以及correlation,其中correlation特性在滤波器设计,信号检测,信号预测以及系统识 ...

  5. Power Spectral Density

    对于一个特定的信号来说,有时域与频域两个表达形式,时域表现的是信号随时间的变化,频域表现的是信号在不同频率上的分量.在信号处理中,通常会对信号进行傅里叶变换得到该信号的频域表示,从而得到信号在频域上的 ...

  6. System.Diagnostics.Process.Star的用法

    System.Diagnostics.Process.Start(); 能做什么呢?它主要有以下几个功能: 1.打开某个链接网址(弹窗). 2.定位打开某个文件目录. 3.打开系统特殊文件夹,如“控制 ...

  7. System.Diagnostics.Process 测试案例

    1.System.Diagnostics.Process 执行exe文件 创建项目,编译成功后,然后把要运行的exe文件拷贝到该项目的运行工作目录下即可,代码如下: using System; usi ...

  8. Unable to extract 64-bitimage. Run Process Explorer from a writeable directory

    Unable to extract 64-bitimage. Run Process Explorer from a writeable directory When we run Process E ...

  9. Linux Process VS Thread VS LWP

    Process program program==code+data; 一个进程可以对应多个程序,一个程序也可以变成多个进程.程序可以作为一种软件资源长期保存,以文件的形式存放在硬盘 process: ...

随机推荐

  1. AQS实现原理分析——ReentrantLock

    在Java并发包java.util.concurrent中可以看到,不少源码是基于AbstractQueuedSynchronizer(以下简写AQS)这个抽象类,因为它是Java并发包的基础工具类, ...

  2. 一起学习造轮子(二):从零开始写一个Redux

    本文是一起学习造轮子系列的第二篇,本篇我们将从零开始写一个小巧完整的Redux,本系列文章将会选取一些前端比较经典的轮子进行源码分析,并且从零开始逐步实现,本系列将会学习Promises/A+,Red ...

  3. JMX,Jstatd做好JVM应用上线的最后一层保障

    目录 我理解的监控 代码异常监控 远程主机配置jmx 修改要启动的java程序启动参数(JVM_OPTS). 修改文件权限 启动jvisualvm 监控服务器上的java程序 jstatd 连接到远程 ...

  4. c++学习之字符串拼接

    在这里,强调这样一个问题: 可以看出,c++中,定义了string类,string 类方便我们进行字符串的一些操作,而不是像C语言中采用字符数组的方式或者指针的方式,通过上面的一些描述,可以发现: s ...

  5. 把玩Alpine linux(二):APK包管理器

    导读 Alpine Linux非常精简,开机内存占用也在二三十兆大,没有拆箱即用,就需要我们自己去做一些了解和配置 Alpine Linux的优劣 优势 Alpine Linux的Docker镜像特点 ...

  6. Mike and palindrome CodeForces - 798A

    题目链接 一个简单的题目,但是却很少有人可以一次AC,比如我就瞎写wa了一次... 写本博算个教训录吧. 题目给出一个字符串,让你严格的改变一个字符使改变后的字符串是一个回文串. 回文串不用解释了.不 ...

  7. Podfile文件用法详解

    https://www.jianshu.com/p/b8b889610b7e 2018.01.09 15:51* 字数 2343 阅读 6263评论 3喜欢 34 前言 iOS开发会经常用到cocoa ...

  8. [2017BUAA软工助教]个人项目小结

    2017BUAA个人项目小结 一.作业链接 http://www.cnblogs.com/jiel/p/7545780.html 二.评分细则 0.注意事项 按时间完成并提交--正常评分 晚交一周以内 ...

  9. MySQL中有关NULL的计算

    mysql> select NULL=NULL; #判断两个NULL是否相等,结果不是1也不是0 +-----------+ | NULL=NULL | +-----------+ | NULL ...

  10. html,css学习实践总结

    网页的布局方式 1.什么是网页的布局方式? 网页的布局方式其实就是指浏览器是如何对网页中的元素进行排版的 1.标准流(文档流/普通流)排版方式 1.1其实浏览器默认的排版方式就是标准流的排版方式 1. ...