ADSP重点习题第二章-第三章（原版书第四章）

第二章

例题2.1.5

Consider the following harmonic process
$$x(n)=cos(0.1\pi n+{\phi }_1)+2sin(1.5n+{\phi }_2).$$
where ${\phi }_1$ and ${\phi }_2$ are IID random variables uniformly distributed in the interval $[0,2\pi ]$. The first component of $x(n)$ is periodic with ${\omega }_1=0.1\pi$ and period equal to 20 while the second component is almost periodic with ${\omega }_1=0.1\pi =1.5$. Thus sequence $x(n)$ is almost periodic.
The mean of $x(n)$ is
μ x ( n ) = E { x ( n ) } = E { c o s ( 0.1 π n + φ 1 ) + 2 s i n ( 1.5 n + φ 2 ) } = 0 \mu_x(n)=E\{ x(n)\} \\ = E\{cos(0.1\pi n+ \varphi_{1})+2sin(1.5n+\varphi_{2})\} \\ = 0 μx​(n)=E{x(n)}=E{cos(0.1πn+φ1​)+2sin(1.5n+φ2​)}=0
and the autocorrelation sequence(using mutual independence between ${\phi }_1$ and ${\phi }_2$) is:
r x ( n 1 , n 2 ) = E { x ( n 1 ) x ∗ ( n 2 ) } = E { c o s ( 0.1 π n 1 + φ 1 ) c o s ( 0.1 π n 2 + φ 1 ) } + E { 2 s i n ( 1.5 n 1 + φ 2 ) 2 c o s ( 1.5 n 2 + φ 2 ) } = 1 2 c o s [ 0.1 π ( n 1 − n 2 ) ] + 2 c o s [ 1.5 ( n 1 − n 2 ) ] r_{x}(n_{1},n_{2})=E\{ x(n_{1})x^{*}(n_{2})\} \\ =E\{cos(0.1\pi n_{1}+\varphi_{1})cos(0.1\pi n_{2}+\varphi_{1})\} \\ +E\{2sin(1.5n_{1}+\varphi_{2})2cos(1.5n_{2}+\varphi_{2})\} \\ =\frac{1}{2}cos[0.1\pi (n_{1}-n_{2})] + 2cos[1.5(n_{1}-n_{2})] rx​(n1​,n2​)=E{x(n1​)x∗(n2​)}=E{cos(0.1πn1​+φ1​)cos(0.1πn2​+φ1​)}+E{2sin(1.5n1​+φ2​)2cos(1.5n2​+φ2​)}=21​cos[0.1π(n1​−n2​)]+2cos[1.5(n1​−n2​)]
or
$$r_x(l)=\frac{1}{2}cos0.1\pi l+2cos1.5l$$
The power spectrum $R_x(e^{jw})$ is given by:
$$\begin{gathered} R_x(e^{jw})=2\pi \delta (\omega +1.5)+\frac{\pi }{2}\delta (\omega +0.1\pi ) \\ +2\pi \delta (\omega -1.5)+\frac{\pi }{2}\delta (\omega -0.1\pi ) \end{gathered}$$
补充：
积化和差，和差化积公式：

习题2.12

A causal LTI system which is described by the difference equation
$$y(n)=\frac{1}{2}y(n-1)+x(n)+\frac{1}{3}x(n-1)$$
is driven by a zero-mean WSS process with autocorrelation $r_x(l)=0.5^{|l|}$.
(a)Determine the PSD and the autocorrelation of the output sequence $y(n)$.
易得：$$\begin{gathered} H(z)=\frac{1+\frac{1}{3}{z}^{-1}}{1-\frac{1}{2}{z}^{-1}},|z|<\frac{1}{2} \\ {R}_x(z)=\frac{\frac{3}{4}}{\frac{5}{4}-\frac{1}{2}(z+z^{-1})},\frac{1}{2}<|z|<2 \end{gathered}$$

(b)Determine the cross-correlation $r_{xy}(l)$ and cross-PSD $R_{xy}(e^{jw})$ between the input and output signals.

知识点：
$$\begin{gathered} R_{xy}(z)=H^{\ast }(\frac{1}{z^{\ast }})R_x(z) \\ {R}_{yx}(z)=H(z)R_x(z) \\ {R}_y(z)=H(z)H^{\ast }(\frac{1}{z^{\ast }})R_x(z) \end{gathered}$$

习题2.13

A WSS process with PSD $R_x(e^{jw})=1/(1.64+1.6cosw)$ is applied to a causal system described by the following difference equation
$$y(n)=0.6y(n-1)+x(n)+1.25x(n-1)$$
Compute (a)the PSD of the output and (b)the cross-PSD $R_{xy}(e^{jw})$ between input and output.

涉及知识点同上题

第三章

习题3.2

Consider a zero-mean random sequence $x(n)$ with PSD
$$R_x(e^{jw})=\frac{5+3cosw}{17+8cosw}$$
(a)Determine the innovations representation of the process $x(n)$

(b)Find the autocorrelation sequence $r_x(l)$.

知识点：
$$R_x(z)={\sigma }_w^{2}H(z)H^{\ast }(\frac{1}{z^{\ast }})$$

习题3.7

Use the Yule-Walker equations to determine the autocorrelation and partial autocorrelation coefficients of the following AR models,assuming that $w(n)$~$WN(0,1)$.
(a)$x(n)=0.5x(n-1)+w(n)$

(b)$x(n)=1.5x(n-1)-0.6x(n-2)+w(n)$


知识点：

尤利-沃克方程：

对于全极点模型，由(4.2.15)和(4.2.16):
$$\begin{gathered} \sum _{k=0}^P{a}_k{r}_h(l-k)=d_0{h}^{\ast }(-l),-\infty <l<\infty \\ \sum _{k=0}^P{a}_k{r}_h(l-k)=0,l>0 \end{gathered}$$
写成矩阵形式，即尤利-沃克(Yule-Walker)方程：
$$\left[\begin{array}{cccc}{r}_h(0)& r_h(1)& \cdots & r_h(P)\\ r_h^{\ast }(1)& r_h(0)& \cdots & r_h(P-1)\\ ⋮& ⋮& \ddots & ⋮\\ r_h^{\ast }(P)& r_h^{\ast }(P-1)& \cdots & r_h(0)\end{array}\right]\left[\begin{array}{c}1\\ a_1\\ ⋮\\ a_p\end{array}\right]=\left[\begin{array}{c}|d_0{|}^2\\ 0\\ ⋮\\ 0\end{array}\right]$$
即为：
$${\mathbf{R}}_{\mathbf{h}}\mathbf{a}=-{\mathbf{r}}_{\mathbf{h}}$$
同时，我们也可以根据输出过程$x(n)$的自相关表示模型参数：
$${\mathbf{R}}_{\mathbf{x}}\mathbf{a}=-{\mathbf{r}}_{\mathbf{x}}$$
其中${\mathbf{r}}_{\mathbf{x}}={\sigma }_w^2{\mathbf{r}}_{\mathbf{h}}$，因为$r_x(l)={\sigma }_w^2{r}_h(l)$

PACS（部分自相关序列的计算）

$$\left[\begin{array}{cccc}1& \rho (1)& \cdots & \rho (m-1)\\ {\rho }^{\ast }(1)& 1& \cdots & ⋮\\ ⋮& ⋮& \ddots & ⋮\\ {\rho }^{\ast }(m-1)& \cdots & {\rho }^{\ast }(1)& 1\end{array}\right]\left[\begin{array}{c}{a}_1^{(m)}\\ a_2^{(m)}\\ ⋮\\ a_m^{(m)}\end{array}\right]=\left[\begin{array}{c}{\rho }^{\ast }(1)\\ {\rho }^{\ast }(2)\\ ⋮\\ {\rho }^{\ast }(m)\end{array}\right]$$,对于$m>P,a_m^{(m)}=0$，我们可以使用序列$a_m^{(m)},m=1,2,\cdots$这一被称为部分自相关的序列去确定全极点模型的阶。$\rho (l)$为归一化的自相关序列，$\rho (l)=\frac{r(l)}{r(0)}$

习题3.11

Consider the following AR(2) models:(i)$x(n)=0.6x(n-1)+0.3x(n-2)+w(n)$ and (ii)$x(n)=0.8x(n-1)-0.5x(n-2)+w(n)$,where$w(n)$~$WGN(0,1)$.
(a)Find the general expression for the normalized autocorrelation sequence $\rho (l)$, and determine ${\sigma }_x^2$.


(b)Plot ${\rho (l)}_0^{15}$ and check if the models exhibit pseudoperiodic behavior.

From these plots, $x(n)=0.6x(n-1)+0.3x(n-2)+w(n)$ does not show any pseudo-periodic behavior, while $x(n)=0.8x(n-1)-0.5x(n-2)+w(n)$ does show pseudo-periodic behavior.
( c)Justify your answer in part(b) by ploting the PSD of the two models.
Clearly, the first model has no fundamental frequency.While the second system does indeed have a large no zero frequency component.
知识点：
二阶模型确定好$r(0),r(1)$的值和$r(l)$与$r(l-1),r(l-2)$的关系式之后，就可以画出整个$r(l),\rho (l)$曲线。

习题3.21

Consider the MA(2) model $x(n)=w(n)-0.1w(n-1)+0.2w(n-2)$.
(a) Is the process $x(n)$ stationary,why?
The process $x(n)$ is a linear combination of a stationary process $w(n)$, therefore $x(n)$ is stationary.
(b) Is the model minimum-phase,why?

( c) Determine the autocorrelation and the partial autocorrelation of the process.



知识点：

低阶极点模型(低阶AP模型)

一阶全极点模型：AP(1)
一个AP(1)模型有传递函数如下：
$$H(z)=\frac{d_0}{1+a{z}^{-1}}$$,如果$-1<a<1$，则$H(z)$是最小相位的。PACS长度为一：$a_1=-\frac{r(1)}{r(0)}=-\rho (1)$
滤波器冲激响应：$h(n)=d_0(-a{)}^{n}u(n)$
自相关序列：$r(l)=r(0)(-a{)}^{|l|},r(0)=\frac{d_0^2}{1-a^2}$
自相关z变换：$R(z)=\frac{d_0^2}{(1+a{z}^{-1})(1+az)}$
频谱：$R(e^{jw})=\frac{d_0^2}{1+2acosw+a^2}$
二阶全极点模型：AP(2)
AP(2)模型系统函数：
$$H(z)=\frac{d_0}{1+a_1{z}^{-1}+a_2{z}^{-2}}=\frac{d_0}{(1-p_1{z}^{-1})(1-p_2{z}^{-1})}$$
其中：$a_1=-(p_1+p_2),a_2=p_1{p}_2$
最小相位条件：
$$\begin{gathered} -1<a_2<1 \\ {a}_2-a_1>-1 \\ {a}_2+a_1>-1 \end{gathered}$$

低阶零点模型(低阶MA模型)

一阶全零点模型：AZ(1)
$$H(z)=G(1+d_1{z}^{-1})$$ $d_1$取任何值，模型稳定，当$-1<d_1<1$时相位最小。
$$R_h(z)=H(z)H(z^{-1})=G^2[d_{1}z+(1+d_1^2)+d_1{z}^{-1}]$$
自相关函数为上式的z反变换，有$r_h(0)=G^2(1+d_1^2),r_h(1)=r_h(-1)=G^2{d}_1$,其他情况时r_h(l)=0。归一化自相关函数：
$$\rho (l)=\{\begin{array}{lr}1& l=0\\ \frac{d_1}{1+d_1^2}& l=\pm 1\\ 0& |l|\ge 2\end{array}$$
二阶全零点模型：AZ(2)
系统函数为：$H(z)=G(1+d_1{z}^{-1}+d_2{z}^{-2})$
对于$d_1,d_2$所有值，系统稳定，如果满足：
$$\begin{gathered} -1<d_2<1 \\ {d}_2-d_1>-1 \\ {d}_2+d_1>-1 \end{gathered}$$
则系统为最小相位系统。归一化自相关函数：
$$\rho (l)=\{\begin{array}{lr}1& l=0\\ \frac{d_1(1+d_2)}{1+d_1^2+d_2^2}& l=\pm 1\\ \frac{d_2}{1+d_1^2+d_2^2}& l=\pm 2\\ 0& |l|\ge 2\end{array}$$
频谱：$R_h(e^{jw})=G^2[(1+d_1^2+d_2^2)+2d_1(1+d_2)cosw+2d_{2}cos2w]$

习题3.23

Determine the coefficients of a PZ(2,1) model with autocorrelation values $r_h(0)=19,r_h(1)=9,r_h(2)=-5,r_h(3)=-7$
易得：
$$\left[\begin{array}{cc}9& 19\\ -5& 9\end{array}\right]\left[\begin{array}{c}{a}_1\\ a_2\end{array}\right]=\left[\begin{array}{c}5\\ 7\end{array}\right]$$
所以，$a_1=-1/2,a_2=1/2$。由
$$r_a(l)=\sum _{k=0}^{P-|l|}{a}_k{a}_{k+|l|}^{\ast },-P\le l\le P$$得：$r_a(0)=3/2,r_a(\pm 1)=-3/4,r_a(\pm 2)=1/2$由：
$$r_d(l)=\sum _{k=-P}^P{r}_a(k)r_h(l-k)$$得：$R_d(z)=4z+10+4z^{-1}=4(1+\frac{1}{2z^{-1}})(z+2)$取其因果部分得到$D(z)$，即$D(z)=2[1+1/(2z^{-1})]$和$d_1=1/2$。
知识点：

极点-零点模型

$$x(n)=-\sum _{k=1}^P{a}_{k}x(n-k)+\sum _{k=0}^Q{d}_{k}w(n-k)$$
其冲激响应可以写成递归形式：
$$h(n)=-\sum _{k=1}^P{a}_{k}h(n-k)+d_n,n\ge 0$$
其中，$d_n=0,n>Q$
自相关：
$H(Z)$的复谱：
$$R_h(z)=H(z)H(\frac{1}{z^{\ast }})=\frac{D(z)D(1/z^{\ast })}{A(z)A(1/z^{\ast })}=\frac{R_d(z)}{R_a(z)}$$
对于自相关序列$r_h(l)$，有：
$$\sum _{k=0}^P{a}_k{r}_h(l-k)=0,l>Q$$,写成矩阵形式为：
$$\left[\begin{array}{cccc}{r}_h(Q)& r_h(Q+1)& \cdots & r_h(Q+P-1)\\ r_h^{\ast }(Q-1)& r_h(Q)& \cdots & r_h(Q+P-2)\\ ⋮& ⋮& \ddots & ⋮\\ r_h^{\ast }(Q-P+1)& r_h^{\ast }(Q-P+2)& \cdots & r_h(Q)\end{array}\right]\left[\begin{array}{c}{a}_1\\ a_2\\ ⋮\\ a_p\end{array}\right]=\left[\begin{array}{c}|r_h(Q-1)\\ r_h(Q-2)\\ ⋮\\ r_h(Q-P)\end{array}\right]$$
同时有：
$$r_a(l)=\sum _{k=0}^{P-|l|}{a}_k{a}_{k+|l|}^{\ast },-P\le l\le P$$因为$R_d(z)=R_a(z)R_h(z)$,所以有：
$$r_d(l)=\sum _{k=-P}^P{r}_a(k)r_h(l-k)$$