第二章 随机信号的噪声
随机过程的基本概念
随机过程的定义
概率分布函数和概率密度函数
一维概率分布函数: F ( x 1 , t 1 ) = P [ X ( t 1 ) ≤ x 1 ] F\left(x_{1}, t_{1}\right)=P\left[X\left(t_{1}\right) \leq x_{1}\right] F(x1,t1)=P[X(t1)≤x1]
一维概率密度函数: ∂ F ( x 1 , t 1 ) ∂ x 1 = f ( x 1 , t 1 ) \frac{\partial F\left(x_{1}, t_{1}\right)}{\partial x_{1}}=f\left(x_{1}, t_{1}\right) ∂x1∂F(x1,t1)=f(x1,t1)
n维概率分布函数: F X ( x 1 , x 2 , ⋯ , x N ; t 1 , t 2 , ⋯ , t N ) = P [ X 1 ⩽ x 1 , X 2 ⩽ x 2 , ⋯ , X N ⩽ x N ] F_{X}\left(x_{1}, x_{2}, \cdots, x_{N} ; t_{1}, t_{2}, \cdots, t_{N}\right)=P\left[X_{1} \leqslant x_{1}, X_{2} \leqslant x_{2}, \cdots, X_{N} \leqslant x_{N}\right] FX(x1,x2,⋯,xN;t1,t2,⋯,tN)=P[X1⩽x1,X2⩽x2,⋯,XN⩽xN]
n维概率密度函数: f X ( x 1 , x 2 , … x n ; t 1 , t 2 … t n ) = ∂ F X ( x 1 , x 2 , … x n ; t 1 , t 2 … t n ) ∂ x 1 , ∂ x 2 … ∂ x n f_{X}\left(x_{1}, x_{2}, \ldots x_{n} ; t_{1}, t_{2} \ldots t_{n}\right)=\frac{\partial F_{X}\left(x_{1}, x_{2}, \ldots x_{n} ; t_{1}, t_{2} \ldots t_{n}\right)}{\partial x_{1}, \partial x_{2} \ldots \partial x_{n}} fX(x1,x2,…xn;t1,t2…tn)=∂x1,∂x2…∂xn∂FX(x1,x2,…xn;t1,t2…tn)
数字特征函数
数学期望(均值函数)
m X ( t ) = E { X ( t ) } ≜ ∫ − ∞ ∞ x f X ( x ; t ) d t m_{X}(t)=E\{X(t)\} \triangleq \int_{-\infty}^{\infty} x f_{X}(x ; t) \mathrm{d} t mX(t)=E{X(t)}≜∫−∞∞xfX(x;t)dt
E { Y ( t ) } = E { g ( X ( t ) ) } ≜ ∫ − ∞ ∞ g ( x ) f X ( x ; t ) d t E\{Y(t)\}=E\{g(X(t))\} \triangleq \int_{-\infty}^{\infty} g(x) f_{X}(x ; t) \mathrm{d} t E{Y(t)}=E{g(X(t))}≜∫−∞∞g(x)fX(x;t)dt
均方函数与方差函数
均方函数
Ψ X 2 ( t ) ≜ E { X 2 ( t ) } = ∫ − ∞ ∞ x 2 f X ( x ; t ) d x \Psi_{X}^{2}(t) \triangleq E\left\{X^{2}(t)\right\}=\int_{-\infty}^{\infty} x^{2} f_{X}(x ; t) \mathrm{d} x ΨX2(t)≜E{X2(t)}=∫−∞∞x2fX(x;t)dx
方差函数:
σ X 2 ( t ) = Var { X ( t ) } = D [ X ( t ) ] ≜ E { [ X ( t ) − m X ( t ) ] 2 } = ∫ − ∞ ∞ [ x − m X ( t ) ] 2 f X ( x ; t ) d x \begin{aligned} \sigma_{X}^{2}(t) &=\operatorname{Var}\{X(t)\}=D[X(t)] \triangleq E\left\{\left[X(t)-m_{X}(t)\right]^{2}\right\} \\ &=\int_{-\infty}^{\infty}\left[x-m_{X}(t)\right]^{2} f_{X}(x ; t) \mathrm{d} x \end{aligned} σX2(t)=Var{X(t)}=D[X(t)]≜E{[X(t)−mX(t)]2}=∫−∞∞[x−mX(t)]2fX(x;t)dx
Var [ X ( t ) ] = Ψ X 2 ( t ) − m X 2 ( t ) \operatorname{Var}[X(t)]=\Psi_{X}^{2}(t)-m_{X}{ }^{2}(t) Var[X(t)]=ΨX2(t)−mX2(t)
相关函数
自相关函数:
R X ( t 1 , t 2 ) ≜ E { X ( t 1 ) X ( t 2 ) } = ∫ − ∞ ∞ ∫ − ∞ ∞ x 1 x 2 f X ( x 1 , x 2 ; t 1 , t 2 ) d x 1 d x 2 R_{X}\left(t_{1}, t_{2}\right) \triangleq E\left\{X\left(t_{1}\right) X\left(t_{2}\right)\right\}=\int_{-\infty}^{\infty} \int_{-\infty}^{\infty} x_{1} x_{2} f_{X}\left(x_{1}, x_{2} ; t_{1}, t_{2}\right) \mathrm{d} x_{1} \mathrm{~d} x_{2} RX(t1,t2)≜E{X(t1)X(t2)}=∫−∞∞∫−∞∞x1x2fX(x1,x2;t1,t2)dx1 dx2
自协方差函数:
C X ( t 1 , t 2 ) ≜ E { [ X ( t 1 ) − m X ( t 1 ) ] [ X ( t 2 ) − m X ( t 2 ) ] } = ∫ − ∞ ∞ ∫ − ∞ ∞ [ x 1 − m X ( t 1 ) ] [ x 2 − m X ( t 2 ) ] f X ( x 1 , x 2 ; t 1 , t 2 ) d x 1 d x 2 = R X ( t 1 , t 2 ) − m X ( t 1 ) m X ( t 2 ) \begin{aligned} & C_{X}\left(t_{1}, t_{2}\right) \triangleq E\left\{\left[X\left(t_{1}\right)-m_{X}\left(t_{1}\right)\right]\left[X\left(t_{2}\right)-m_{X}\left(t_{2}\right)\right]\right\} \\ =& \int_{-\infty}^{\infty} \int_{-\infty}^{\infty}\left[x_{1}-m_{X}\left(t_{1}\right)\right]\left[x_{2}-m_{X}\left(t_{2}\right)\right] f_{X}\left(x_{1}, x_{2} ; t_{1}, t_{2}\right) \mathrm{d} x_{1} \mathrm{~d} x_{2}\\=&R_{X}\left(t_{1}, t_{2}\right)-m_{X}\left(t_{1}\right) m_{X}\left(t_{2}\right) \end{aligned} ==CX(t1,t2)≜E{[X(t1)−mX(t1)][X(t2)−mX(t2)]}∫−∞∞∫−∞∞[x1−mX(t1)][x2−mX(t2)]fX(x1,x2;t1,t2)dx1 dx2RX(t1,t2)−mX(t1)mX(t2)
讨论:
- 当 m X ( t 1 ) = m X ( t 2 ) = 0 m_X(t_1)=m_X(t_2)=0 mX(t1)=mX(t2)=0时, C X ( t 1 , t 2 ) = R X ( t 1 , t 2 ) C_X(t_1,t_2)=R_X(t_1,t_2) CX(t1,t2)=RX(t1,t2)
- 当 t 1 = t 2 = t t_1=t_2=t t1=t2=t 时:此时, 自相关函数是均方函数, 协方差函数就是方差函数
R X ( t , t ) = E { X 2 ( t ) } = Ψ X 2 ( t ) R_{X}(t, t)=E\left\{X^{2}(t)\right\}=\Psi_{X}^{2}(t) RX(t,t)=E{X2(t)}=ΨX2(t)
C X ( t , t ) = E { [ X ( t ) − m X ( t ) ] 2 } = Var [ X ( t ) ] = σ x 2 ( t ) C_{X}(t, t)=E\left\{\left[X(t)-m_{X}(t)\right]^{2}\right\}=\operatorname{Var}[X(t)]=\sigma_{x}^{2}(t) CX(t,t)=E{[X(t)−mX(t)]2}=Var[X(t)]=σx2(t)
Var [ X ( t ) ] = Ψ X 2 ( t ) − m X 2 ( t ) \operatorname{Var}[X(t)]=\Psi_{X}^{2}(t)-m_{X}^{2}(t) Var[X(t)]=ΨX2(t)−mX2(t)
互相关函数
R X Y ( t 1 , t 2 ) ≜ E { X ( t 1 ) Y ( t 2 ) } = ∫ − ∞ x 1 ∫ − ∞ y 2 x 1 y 2 f X Y ( x 1 , t 1 ; y 2 , t 2 ) d x 1 d y 2 R_{X Y}\left(t_{1}, t_{2}\right) \triangleq E\left\{X\left(t_{1}\right) Y\left(t_{2}\right)\right\}=\int_{-\infty}^{x_{1}} \int_{-\infty}^{y_{2}} x_{1} y_{2} f_{X Y}\left(x_{1}, t_{1} ; y_{2}, t_{2}\right) \mathrm{d} x_{1} \mathrm{~d} y_{2} RXY(t1,t2)≜E{X(t1)Y(t2)}=∫−∞x1∫−∞y2x1y2fXY(x1,t1;y2,t2)dx1 dy2
互协方差函数
C X Y ( t 1 , t 2 ) ≜ E { [ X ( t 1 ) − m X ( t 1 ) ] [ Y ( t 2 ) − m Y ( t 2 ) ] } = R X Y ( t 1 , t 2 ) − m X ( t 1 ) m Y ( t 2 ) C_{X Y}\left(t_{1}, t_{2}\right) \triangleq E\left\{\left[X\left(t_{1}\right)-m_{X}\left(t_{1}\right)\right]\left[Y\left(t_{2}\right)-m_{Y}\left(t_{2}\right)\right]\right\}=R_{X Y}\left(t_{1}, t_{2}\right)-m_{X}\left(t_{1}\right) m_{Y}\left(t_{2}\right) CXY(t1,t2)≜E{[X(t1)−mX(t1)][Y(t2)−mY(t2)]}=RXY(t1,t2)−mX(t1)mY(t2)
讨论相互关系:
- 互不相关: C X Y ( t 1 , t 2 ) = 0 C_{X Y}\left(t_{1}, t_{2}\right)=0 CXY(t1,t2)=0,或 R X Y ( t 1 , t 2 ) = m X ( t 1 ) m Y ( t 2 ) R_{X Y}\left(t_{1}, t_{2}\right)=m_{X}\left(t_{1}\right) m_{Y}\left(t_{2}\right) RXY(t1,t2)=mX(t1)mY(t2)
- 相互正交: R X Y ( t 1 , t 2 ) = 0 R_{X Y}\left(t_{1}, t_{2}\right)=0 RXY(t1,t2)=0
- 相互独立: f X Y ( x 1 , t 1 ; y 2 , t 2 ) = f X ( x 1 , t 1 ) f Y ( y 2 , t 2 ) f_{X Y}\left(x_{1}, t_{1} ; y_{2}, t_{2}\right)=f_{X}\left(x_{1}, t_{1}\right) f_{Y}\left(y_{2}, t_{2}\right) fXY(x1,t1;y2,t2)=fX(x1,t1)fY(y2,t2)
- 均值为0,不相关=正交
- 相互独立必不相关,反之不成立
随机过程分类
- 马尔可夫过程:当已知当前状态,则其将来状态与过去状态无关(无后效性)
- 平稳随机过程:过程的统计特性与时间的起点无关,狭义平稳
f X ( x 1 , x 2 , ⋯ , x n ; t 1 , t 2 , ⋯ , t n ) = f X ( x 1 , x 2 , ⋯ , x n ; t 1 + ε , t 2 + ε , ⋯ , t n + ε ) f_{X}\left(x_{1}, x_{2}, \cdots, x_{n} ; t_{1}, t_{2}, \cdots, t_{n}\right)=f_{X}\left(x_{1}, x_{2}, \cdots, x_{n} ; t_{1}+\varepsilon, t_{2}+\varepsilon, \cdots, t_{n}+\varepsilon\right) fX(x1,x2,⋯,xn;t1,t2,⋯,tn)=fX(x1,x2,⋯,xn;t1+ε,t2+ε,⋯,tn+ε)
广义平稳过程及各态遍历性
广义平稳判定:
E { X ( t ) } = ∫ − ∞ ∞ x f X ( x ) d x = m X R X ( t 1 , t 2 ) = ∫ − ∞ ∞ ∫ − ∞ ∞ x 1 x 2 f X ( x 1 , x 2 , t 2 − t 1 ) d x 1 d x 2 = R X ( t 2 − t 1 ) = R X ( τ ) , τ = t 2 − t 1 \begin{aligned} E\{X(t)\} &=\int_{-\infty}^{\infty} x f_{X}(x) \mathrm{d} x=m_{X} \\ R_{X}\left(t_{1}, t_{2}\right) &=\int_{-\infty}^{\infty} \int_{-\infty}^{\infty} x_{1} x_{2} f_{X}\left(x_{1}, x_{2}, t_{2}-t_{1}\right) \mathrm{d} x_{1} \mathrm{~d} x_{2} \\ &=R_{X}\left(t_{2}-t_{1}\right)=R_{X}(\tau), \quad \tau=t_{2}-t_{1} \end{aligned} E{X(t)}RX(t1,t2)=∫−∞∞xfX(x)dx=mX=∫−∞∞∫−∞∞x1x2fX(x1,x2,t2−t1)dx1 dx2=RX(t2−t1)=RX(τ),τ=t2−t1
广义平稳过程相关函数的性质
广义平稳实随机过程的自相关函数性质
-
R X ( 0 ) = E { X 2 ( t ) } ⩾ 0 R_{X}(0)=E\left\{X^{2}(t)\right\} \geqslant 0 RX(0)=E{X2(t)}⩾0
-
R X ( τ ) = R X ( − τ ) R_{X}(\tau)=R_{X}(-\tau) RX(τ)=RX(−τ)
-
R X ( 0 ) ⩾ ∣ R X ( τ ) ∣ R_{X}(0) \geqslant\left|R_{X}(\tau)\right| RX(0)⩾∣RX(τ)∣
-
如果 E { X ( t ) } − m X ( t ) ≠ 0 E\{X(t)\}-m_X(t)≠0 E{X(t)}−mX(t)=0,那么 R X ( τ ) R_X(\tau) RX(τ)将含有常数项 m X m_X mX。满足: lim r → ∞ R X ( τ ) = lim r → ∞ E { X ( t ) X ( t + τ ) } = E { X ( t ) } E { X ( t + τ ) } = m X 2 \begin{aligned} \lim _{r \rightarrow \infty} R_{X}(\tau) &=\lim _{r \rightarrow \infty} E\{X(t) X(t+\tau)\} \\ &=E\{X(t)\} E\{X(t+\tau)\}=m_{X}^{2} \end{aligned} r→∞limRX(τ)=r→∞limE{X(t)X(t+τ)}=E{X(t)}E{X(t+τ)}=mX2
-
如果 X ( t ) X(t) X(t)含有周期性分量,那么 R X ( τ ) R_X (\tau) RX(τ)也将含有相同周期的周期性分量。
-
R X ( τ ) R_X (\tau) RX(τ)不能取随意的形状。
广义平稳实随机过程的互相关函数性质
- R X Y ( − τ ) = R Y X ( τ ) R_{X Y}(-\tau)=R_{Y X}(\tau) RXY(−τ)=RYX(τ)
- ∣ R X Y ( τ ) ∣ ⩽ R X ( 0 ) R Y ( 0 ) \left|R_{X Y}(\tau)\right| \leqslant \sqrt{R_{X}(0) R_{Y}(0)} ∣RXY(τ)∣⩽RX(0)RY(0)
- ∣ R X Y ( τ ) ∣ ⩽ 1 2 [ R X ( 0 ) + R Y ( 0 ) ] \left|R_{X Y}(\tau)\right| \leqslant \frac{1}{2}\left[R_{X}(0)+R_{Y}(0)\right] ∣RXY(τ)∣⩽21[RX(0)+RY(0)]
时间平均及各态遍历性
时间平均
-
时间平均: ⟨ X ( t ) ⟩ = lim T → ∞ 1 2 T ∫ − T T X ( t ) d t \langle X(t)\rangle=\lim _{T \rightarrow \infty} \frac{1}{2 T} \int_{-T}^{T} X(t) \mathrm{d} t ⟨X(t)⟩=limT→∞2T1∫−TTX(t)dt
-
时间平均自相关函数:
⟨ X ( t ) X ( t + τ ) ⟩ = lim T → ∞ 1 2 T ∫ − T T X ( t ) X ( t + τ ) d t \langle X(t) X(t+\tau)\rangle=\lim _{T \rightarrow \infty} \frac{1}{2 T} \int_{-T}^{T} X(t) X(t+\tau) \mathrm{d} t ⟨X(t)X(t+τ)⟩=limT→∞2T1∫−TTX(t)X(t+τ)dt
各态遍历性
各态遍历性: 时间均值和时间自相关函数分别等于统计平均值和统计自相关函数。这个性质称 为各态遍历性, 而满足这个性质的平稳过程称为遍历过程。
E { ⟨ X ( t ) ⟩ } = E { X ( t ) } = m X E { ⟨ X ( t ) X ( t + τ ) ⟩ } = R X ( τ ) \begin{array}{l} E\{\langle X(t)\rangle\}=E\{X(t)\}=m_{X} \\ E\{\langle X(t) X(t+\tau)\rangle\}=R_{X}(\tau) \end{array} E{⟨X(t)⟩}=E{X(t)}=mXE{⟨X(t)X(t+τ)⟩}=RX(τ)
对于大多数平稳过程, 随机变量 ⟨ X ( t ) ⟩ \langle X(t)\rangle ⟨X(t)⟩ 和 ⟨ X ( t ) X ( t + τ ) ⟩ \langle X(t) X(t+\tau)\rangle ⟨X(t)X(t+τ)⟩的方差随着 T → ∞ T \rightarrow \infty T→∞而趋于 0, 那么,只需
⟨ X ( t ) ⟩ = E { X ( t ) } ⟨ X ( t ) X ( t + τ ) ⟩ = R X ( τ ) \begin{array}{l} \langle X(t)\rangle=E\{X(t)\} \\ \langle X(t) X(t+\tau)\rangle=R_{X}(\tau) \end{array} ⟨X(t)⟩=E{X(t)}⟨X(t)X(t+τ)⟩=RX(τ)
也就是说, 各态遍历定理指出, 对于非周期平稳过程 X ( t ) X(t) X(t) , 只要
lim τ → ∞ C X ( τ ) = 0 \lim _{\tau \rightarrow \infty} C_{X}(\tau)=0 limτ→∞CX(τ)=0,那么过程 X ( t ) X(t) X(t)是均值遍历的。
随机信号 X ( t ) X(t) X(t)可测量的参数如下:
- E { X ( t ) } = ⟨ X ( t ) ⟩ E\{X(t)\}=\langle X(t)\rangle E{X(t)}=⟨X(t)⟩ 随机信号的直流分量;
- E { X 2 ( t ) } = ⟨ X 2 ( t ) ⟩ E\left\{X^{2}(t)\right\}=\left\langle X^{2}(t)\right\rangle E{X2(t)}=⟨X2(t)⟩ 随机信号的平均功率;
- R X ( τ ) = ⟨ X ( t ) X ( t + τ ) ⟩ ≈ 1 T ∫ t H T X ( t ) X ( t − τ ) d t R_{X}(\tau)=\langle X(t) X(t+\tau)\rangle \approx \frac{1}{T} \int_{t}^{H T} X(t) X(t-\tau) \mathrm{d} t RX(τ)=⟨X(t)X(t+τ)⟩≈T1∫tHTX(t)X(t−τ)dt
随机过程的功率谱密度
- Parseval定理: ∫ − ∞ ∞ x 2 ( t ) d t = ∫ − ∞ ∞ ∣ X ( f ) ∣ 2 d f \int_{-\infty}^{\infty} x^{2}(t) \mathrm{d} t=\int_{-\infty}^{\infty}|X(f)|^{2} \mathrm{~d} f ∫−∞∞x2(t)dt=∫−∞∞∣X(f)∣2 df, ∣ X ( f ) ∣ 2 |X(f)|^{2} ∣X(f)∣2——能谱密度
- 归一化能量: E = ∫ − ∞ ∞ x 2 ( t ) d t < ∞ E=\int_{-\infty}^{\infty} x^{2}(t) \mathrm{d} t<\infty E=∫−∞∞x2(t)dt<∞, X ( t ) X(t) X(t)能量有限,称为能量信号
- 傅里叶变换: X ( f ) = ∫ − ∞ ∞ x ( t ) e − j 2 π f t d t X(f)=\int_{-\infty}^{\infty} x(t) \mathrm{e}^{-\mathrm{j} 2 \pi f t} \mathrm{~d} t X(f)=∫−∞∞x(t)e−j2πft dt
功率谱密度的定义
- 功率谱密度: S X ( f ) = lim T → ∞ E { ∣ X T ( f ) ∣ 2 } 2 T S_{X}(f)=\lim _{T \rightarrow \infty} \frac{E\left\{\left|X_{\mathrm{T}}(f)\right|^{2}\right\}}{2 T} SX(f)=limT→∞2TE{∣XT(f)∣2},该定义既包含时间平均又包含统计平均
- 归一化功率: P = ∫ − ∞ ∞ S X ( f ) d f P=\int_{-\infty}^{\infty} S_X(f) \mathrm{d} f P=∫−∞∞SX(f)df
求解平均功率的两种方法:
方法1:通过功率谱密度求解, P = ∫ − ∞ ∞ S X ( f ) d f P=\int_{-\infty}^{\infty} S_X(f) \mathrm{d} f P=∫−∞∞SX(f)df
方法2:通过互相关函数求解, P X = R X ( 0 ) = E [ X 2 ( t ) ] P_X=R_X(0)=E[X^2(t)] PX=RX(0)=E[X2(t)]
功率谱密度和自相关函数的关系
-
一般情况: ∫ − ∞ ∞ S X ( f ) e j 2 x f d f = ⟨ R X ( t , t + τ ) ⟩ \int_{-\infty}^{\infty} S_{X}(f) \mathrm{e}^{\mathrm{j} 2 \mathrm{xf}} \mathrm{d} f=\left\langle R_{X}(t, t+\tau)\right\rangle ∫−∞∞SX(f)ej2xfdf=⟨RX(t,t+τ)⟩
-
对于广义平稳的特殊情况,因为 ⟨ R X ( t , t + τ ) ⟩ = R X ( τ ) \left\langle R_{X}(t, t+\tau)\right\rangle=R_{X}(\tau) ⟨RX(t,t+τ)⟩=RX(τ),于是有:
维纳-辛钦定理: S X ( f ) = ∫ − ∞ ∞ R X ( τ ) e − j 2 π f τ d τ R X ( τ ) = ∫ − ∞ ∞ S X ( f ) e j 2 π f τ d f \begin{array}{l} S_{X}(f)=\int_{-\infty}^{\infty} R_{X}(\tau) \mathrm{e}^{-\mathrm{j} 2 \pi f \tau} \mathrm{d} \tau \\ R_{X}(\tau)=\int_{-\infty}^{\infty} S_{X}(f) \mathrm{e}^{\mathrm{j} 2 \pi f \tau}\mathrm{d} f \end{array} SX(f)=∫−∞∞RX(τ)e−j2πfτdτRX(τ)=∫−∞∞SX(f)ej2πfτdf
功率谱密度的性质
- S X ( f ) ≥ 0 S_X(f)\ge0 SX(f)≥0
- S X ( − f ) = S X ( f ) S_X(-f)=S_X(f) SX(−f)=SX(f),如 X ( t ) X_(t) X(t)为实且偶的函数
- S X ( f ) S_X(f) SX(f)总是实函数
其他:
-
相关时间: τ 0 = 1 2 ⋅ S X ( 0 ) C X ( 0 ) \tau_0=\frac{1}{2}·\frac{S_X(0)}{C_X(0)} τ0=21⋅CX(0)SX(0)
反映随机过程自相关性的强弱: τ 0 \tau_0 τ0越小, C X ( τ ) C_X(\tau) CX(τ)越陡,相关程度越弱;时域波形起伏越大,含有丰富的高频分量,等效带宽越大,相似度越低
-
等效带宽: B N = 1 2 ⋅ C X ( 0 ) S X ( 0 ) B_N=\frac{1}{2}·\frac{C_X(0)}{S_X(0)} BN=21⋅SX(0)CX(0)
与 τ 0 \tau_0 τ0数值上成反比
互功率谱密度及其性质
互功率谱密度定义: S X 1 X 2 ( f ) = lim T → ∞ E { X 1 T ∗ ( f ) X 2 T ( f ) } 2 T S_{X_1 X_2}(f)=\lim _{T \rightarrow \infty} \frac{E\left\{X_{1\mathrm{T}}^*(f)X_{2\mathrm{T}}(f) \right\}}{2 T} SX1X2(f)=limT→∞2TE{X1T∗(f)X2T(f)}
对于联合平稳过程 X ( t ) X(t) X(t) 和 Y ( t ) Y(t) Y(t) , 我们直接定义它们的互功率谱密度 (以下简称互谱密 度)为互相关函数的傅里叶变换及其逆变换:
S X Y ( f ) ≜ ∫ − ∞ ∞ R X Y ( τ ) e − j ω τ d τ S_{X Y}(f) \triangleq \int_{-\infty}^{\infty} R_{X Y}(\tau) \mathrm{e}^{-\mathrm{j} \omega \tau} \mathrm{d} \tau SXY(f)≜∫−∞∞RXY(τ)e−jωτdτ
R X Y ( τ ) = ∫ − ∞ ∞ S X Y ( f ) e j ω τ d f R_{X Y}(\tau)=\int_{-\infty}^{\infty} S_{X Y}(f) \mathrm{e}^{\mathrm{j} \omega \tau} \mathrm{d} f RXY(τ)=∫−∞∞SXY(f)ejωτdf
互谱密度有如下性质: S X Y ( f ) = S Y X ∗ ( f ) = S Y X ( − f ) S_{X Y}(f)=S_{Y X}^{*}(f)=S_{Y X}(-f) SXY(f)=SYX∗(f)=SYX(−f),因此互谱密度没有自功率谱那样的对称性, 并且通常是复的函数。
线性系统对随机信号的响应
输出随机信号: Y ( t ) = ∫ − ∞ ∞ h ( τ ) X ( t − τ ) d τ Y(t)=\int_{-\infty}^{\infty}h(\tau)X(t-\tau)\mathrm{d}\tau Y(t)=∫−∞∞h(τ)X(t−τ)dτ
频域表示: Y ( f ) = H ( f ) X ( f ) Y(f)=H(f)X(f) Y(f)=H(f)X(f)
输出过程的统计特性
输出过程的均值
E { Y ( t ) } = ∫ − ∞ ∞ E { X ( t − τ ) } h ( τ ) d τ E\{Y(t)\}=\int_{-\infty}^{\infty} E\{X(t-\tau)\} h(\tau) \mathrm{d} \tau E{Y(t)}=∫−∞∞E{X(t−τ)}h(τ)dτ
当 X ( t ) X(t) X(t)是广义平稳过程时, E { Y ( t ) } = E { X ( t ) } ∫ − ∞ ∞ h ( τ ) d τ = m X H ( 0 ) E\{Y(t)\}=E\{X(t)\} \int_{-\infty}^{\infty} h(\tau) \mathrm{d} \tau=m_{X} H(0) E{Y(t)}=E{X(t)}∫−∞∞h(τ)dτ=mXH(0)
输出过程的自相关函数
R y ( t 1 , t 2 ) = E [ Y ( t 1 ) Y ( t 2 ) ] = E { ∫ − ∞ ∞ X ( t 1 − u ) h ( u ) d u ∫ − ∞ ∞ X ( t 2 − v ) h ( v ) d v } = ∫ − ∞ ∞ ∫ − ∞ ∞ E { X ( t 1 − u ) X ( t 2 − v ) } h ( u ) h ( v ) d u d v = ∫ − ∞ ∞ ∫ − ∞ ∞ R X ( t 1 − u , t 2 − v ) h ( u ) h ( v ) d u d v \begin{aligned} R_{y}\left(t_{1}, t_{2}\right) &=E\left[Y\left(t_{1}\right) Y\left(t_{2}\right)\right] \\ &=E\left\{\int_{-\infty}^{\infty} X\left(t_{1}-u\right) h(u) \mathrm{d} u \int_{-\infty}^{\infty} X\left(t_{2}-v\right) h(v) \mathrm{d} v\right\} \\ &=\int_{-\infty}^{\infty} \int_{-\infty}^{\infty} E\left\{X\left(t_{1}-u\right) X\left(t_{2}-v\right)\right\} h(u) h(v) \mathrm{d} u \mathrm{~d} v \\ &=\int_{-\infty}^{\infty} \int_{-\infty}^{\infty} R_{X}\left(t_{1}-u, t_{2}-v\right) h(u) h(v) \mathrm{d} u \mathrm{~d} v \end{aligned} Ry(t1,t2)=E[Y(t1)Y(t2)]=E{∫−∞∞X(t1−u)h(u)du∫−∞∞X(t2−v)h(v)dv}=∫−∞∞∫−∞∞E{X(t1−u)X(t2−v)}h(u)h(v)du dv=∫−∞∞∫−∞∞RX(t1−u,t2−v)h(u)h(v)du dv
当 X ( t ) X(t) X(t) 是广义平稳过程时,输出过程 Y ( t ) Y(t) Y(t) 也是广义平稳的,令 $ t_{2}-t_{1}=\tau$,则
R Y ( τ ) = ∫ − ∞ ∞ ∫ − ∞ ∞ R X ( τ + u − v ) h ( u ) h ( v ) d u d v = R X ( τ ) ∗ h ( τ ) ∗ h ( − τ ) = R X ( τ ) ∗ R h ( τ ) \begin{aligned} R_{Y}(\tau) &=\int_{-\infty}^{\infty} \int_{-\infty}^{\infty} R_{X}(\tau+u-v) h(u) h(v) \mathrm{d} u \mathrm{~d} v \\ &=R_{X}(\tau) * h(\tau) * h(-\tau)\\&=R_{X}(\tau) *R_h(\tau) \end{aligned} RY(τ)=∫−∞∞∫−∞∞RX(τ+u−v)h(u)h(v)du dv=RX(τ)∗h(τ)∗h(−τ)=RX(τ)∗Rh(τ)
线性系统脉冲响应的自相关积分:
R h ( τ ) ≜ ∫ − ∞ ∞ h ( τ + l ) h ( l ) d l = ∫ − ∞ ∞ h ( τ − l ′ ) h ( − l ′ ) d l ′ = h ( τ ) ∗ h ( − τ ) \begin{aligned} R_{h}(\tau) & \triangleq \int_{-\infty}^{\infty} h(\tau+l) h(l) \mathrm{d} l=\int_{-\infty}^{\infty} h\left(\tau-l^{\prime}\right) h\left(-l^{\prime}\right) \mathrm{d} l^{\prime} \\ &=h(\tau) * h(-\tau) \end{aligned} Rh(τ)≜∫−∞∞h(τ+l)h(l)dl=∫−∞∞h(τ−l′)h(−l′)dl′=h(τ)∗h(−τ)
输出过程的功率谱
我们可称 ∣ H ( f ) ∣ 2 |H(f)|^{2} ∣H(f)∣2为系统的功率谱传输函数。显然, 它是非负的、实的和偶的函数, 而传输函数 H ( f ) H(f) H(f) 的相位特性对功率谱传输不起作用。
S Y ( f ) = S X ( f ) H ( f ) H ∗ ( f ) = ∣ H ( f ) ∣ 2 S X ( f ) \begin{aligned} S_{Y}(f) &=S_{X}(f) H(f) H^{*}(f) \\ &=|H(f)|^{2} S_{X}(f) \end{aligned} SY(f)=SX(f)H(f)H∗(f)=∣H(f)∣2SX(f)
拓展:
平稳输入通过线性滤波器后,输出过程的互功率谱密度: S Y 1 Y 2 ( f ) = S X ( f ) H 1 ( f ) H 2 ∗ ( f ) S_{Y_1Y_2}(f) =S_{X}(f) H_1(f) H_2^{*}(f) SY1Y2(f)=SX(f)H1(f)H2∗(f)
证明:
方法1:通过互相关函数定义证明,最后对整体进行傅里叶变换即可; R Y 1 Y 2 ( t 1 , t 2 ) ≜ E { Y 1 ( t 1 ) Y 2 ( t 2 ) } = ∫ − ∞ y 1 ∫ − ∞ y 2 y 1 y 2 f Y 1 Y 2 ( y 1 , t 1 ; y 2 , t 2 ) d y 1 d y 2 R_{Y_1 Y_2}\left(t_{1}, t_{2}\right) \triangleq E\left\{Y_1\left(t_{1}\right) Y_2\left(t_{2}\right)\right\}=\int_{-\infty}^{y_{1}} \int_{-\infty}^{y_{2}} y_{1} y_{2} f_{Y_1 Y_2}\left(y_{1}, t_{1} ; y_{2}, t_{2}\right) \mathrm{d} y_{1} \mathrm{~d} y_{2} RY1Y2(t1,t2)≜E{Y1(t1)Y2(t2)}=∫−∞y1∫−∞y2y1y2fY1Y2(y1,t1;y2,t2)dy1 dy2
方法2:通过互功率谱密度定义证明, S X 1 X 2 ( f ) = lim T → ∞ E { X 1 T ∗ ( f ) X 2 T ( f ) } 2 T S_{X_1 X_2}(f)=\lim _{T \rightarrow \infty} \frac{E\left\{X_{1\mathrm{T}}^*(f)X_{2\mathrm{T}}(f) \right\}}{2 T} SX1X2(f)=limT→∞2TE{X1T∗(f)X2T(f)}
输出输入过程相互关系
输入输出过程的互相关函数
R X Y ( t 1 , t 2 ) = E { X ( t 1 ) Y ( t 2 ) } = E { X ( t 1 ) ∫ − ∞ ∞ X ( t 2 − v ) h ( v ) d v } = ∫ − ∞ ∞ R X ( t 1 , t 2 − v ) h ( v ) d v \begin{aligned} R_{X Y}\left(t_{1}, t_{2}\right) &=E\left\{X\left(t_{1}\right) Y\left(t_{2}\right)\right\} \\ &=E\left\{X\left(t_{1}\right) \int_{-\infty}^{\infty} X\left(t_{2}-v\right) h(v) \mathrm{d} v\right\} \\ &=\int_{-\infty}^{\infty} R_{X}\left(t_{1}, t_{2}-v\right) h(v) \mathrm{d} v \end{aligned} RXY(t1,t2)=E{X(t1)Y(t2)}=E{X(t1)∫−∞∞X(t2−v)h(v)dv}=∫−∞∞RX(t1,t2−v)h(v)dv
当 X ( t ) X(t) X(t) 为广义平稳过程时, Y ( t ) Y(t) Y(t)也为广义平稳过程, 并且 X ( t ) X(t) X(t)与 Y ( t ) Y(t) Y(t)是联合广义平 稳的。 t 2 − t 1 = τ t_{2}-t_{1}=\tau t2−t1=τ,则
R X Y ( τ ) = ∫ − ∞ ∞ R X ( τ − v ) h ( v ) d v = R X ( τ ) ∗ h ( τ ) \begin{aligned} R_{X Y}(\tau) &=\int_{-\infty}^{\infty} R_{X}(\tau-v) h(v) \mathrm{d} v \\ &=R_{X}(\tau) * h(\tau) \end{aligned} RXY(τ)=∫−∞∞RX(τ−v)h(v)dv=RX(τ)∗h(τ)
R Y X ( τ ) = ∫ − ∞ ∞ R X ( τ + v ) h ( v ) d v = R X ( τ ) ∗ h ( − τ ) \begin{aligned} R_{Y X}(\tau) &=\int_{-\infty}^{\infty} R_{X}(\tau+v) h(v) \mathrm{d} v \\ &=R_{X}(\tau) * h(-\tau) \end{aligned} RYX(τ)=∫−∞∞RX(τ+v)h(v)dv=RX(τ)∗h(−τ)
输入输出过程的功率谱密度
S X Y ( f ) = H ( f ) S X ( f ) S Y X ( f ) = S X ( f ) H ∗ ( f ) = S X Y ∗ ( f ) \begin{array}{l} S_{X Y}(f)=H(f) S_{X}(f) \\ S_{Y X}(f)=S_{X}(f) H^{*}(f)=S_{X Y}^{*}(f) \end{array} SXY(f)=H(f)SX(f)SYX(f)=SX(f)H∗(f)=SXY∗(f)
高斯过程和白噪声过程
高斯过程
N 维高斯联合概率密度函数为:
f X ( X ) = 1 ( 2 π ) N / 2 ∣ C ∣ 1 / 2 exp { − 1 2 ( X − m ) T C − 1 ( X − m ) } f_{X}(\boldsymbol{X})=\frac{1}{(2 \pi)^{N / 2}|\boldsymbol{C}|^{1 / 2}} \exp \left\{-\frac{1}{2}(\boldsymbol{X}-\boldsymbol{m})^{\mathrm{T}} \boldsymbol{C}^{-1}(\boldsymbol{X}-\boldsymbol{m})\right\} fX(X)=(2π)N/2∣C∣1/21exp{−21(X−m)TC−1(X−m)}
其中:高斯过程的 X ( t ) X(t) X(t)的N维随机变量记为, X = ( X 1 X 2 ⋮ X N ) = ( X ( t 1 ) X ( t 2 ) ⋮ X ( t N ) ) \boldsymbol{X}=\left(\begin{array}{c} X_{1} \\ X_{2} \\ \vdots \\ X_{N} \end{array}\right)=\left(\begin{array}{c} X\left(t_{1}\right) \\ X\left(t_{2}\right) \\ \vdots \\ X\left(t_{N}\right) \end{array}\right) X=⎝⎜⎜⎜⎛X1X2⋮XN⎠⎟⎟⎟⎞=⎝⎜⎜⎜⎛X(t1)X(t2)⋮X(tN)⎠⎟⎟⎟⎞
相应的均值矢量为: m = E { X } = ( m 1 m 2 ⋮ m N ) \boldsymbol{m}=E\{\boldsymbol{X}\}=\left(\begin{array}{c} m_{1} \\ m_{2} \\ \vdots \\ m_{N} \end{array}\right) m=E{X}=⎝⎜⎜⎜⎛m1m2⋮mN⎠⎟⎟⎟⎞
式中 C \boldsymbol{C} C 为 N 维高斯随机变量的协方差矩阵, 其定义式为, C = ( C 11 C 12 ⋯ C 1 N C 21 C 22 ⋯ C 2 N ⋮ ⋮ ⋮ C N 1 C N 2 ⋯ C N N ) \boldsymbol{C}=\left(\begin{array}{cccc} C_{11} & C_{12} & \cdots & C_{1 N} \\ C_{21} & C_{22} & \cdots & C_{2 N} \\ \vdots & \vdots & & \vdots \\ C_{N 1} & C_{N 2} & \cdots & C_{N N} \end{array}\right) C=⎝⎜⎜⎜⎛C11C21⋮CN1C12C22⋮CN2⋯⋯⋯C1NC2N⋮CNN⎠⎟⎟⎟⎞
矩阵元素, C i j = E { ( X i − m i ) ( X j − m j ) } C_{i j}=E\left\{\left(X_{i}-m_{i}\right)\left(X_{j}-m_{j}\right)\right\} Cij=E{(Xi−mi)(Xj−mj)}
对于广义平稳高斯过程, m i = m j = m X C i j = R X ( t j − t i ) − m X 2 = C j i C i i = R X ( 0 ) − m X 2 = σ X 2 \begin{array}{l} m_{i}=m_{j}=m_{X} \\ C_{i j}=R_{X}\left(t_{j}-t_{i}\right)-m_{X}^{2}=C_{j i} \\ C_{ii}=R_{X}(0)-m_{X}^{2}=\sigma_{X}^{2} \end{array} mi=mj=mXCij=RX(tj−ti)−mX2=CjiCii=RX(0)−mX2=σX2
如果 N 个随机变量 $X_{i} $还是各不相关的, C i j = 0 , i ≠ j C_{i j}=0, i \neq j Cij=0,i=j。那么协方差简化为一个对角阵, C = ( σ X 2 0 ⋯ 0 0 σ X 2 ⋯ 0 ⋮ ⋮ ⋱ ⋮ 0 0 ⋯ σ X 2 ) \boldsymbol{C}=\left(\begin{array}{cccc} \sigma_{X}^{2} & 0 & \cdots & 0 \\ 0 & \sigma_{X}^{2} & \cdots & 0 \\ \vdots & \vdots & \ddots & \vdots \\ 0 & 0 & \cdots & \sigma_{X}^{2} \end{array}\right) C=⎝⎜⎜⎜⎛σX20⋮00σX2⋮0⋯⋯⋱⋯00⋮σX2⎠⎟⎟⎟⎞
这时, N 维高斯联合概率密度简化为: f X ( X ) = ∏ i = 1 N f X ( x i ) f_{X}(\boldsymbol{X})=\prod_{i=1}^{N} f_{X}\left(x_{i}\right) fX(X)=∏i=1NfX(xi)
笔记:
- 高斯过程通过线性系统后的输出仍为高斯过程,但数字特征可能变化。推广:线性系统的输入与输出过程是联合高斯的。
- 由于高斯过程N维联合概率分布完全取决于一维和二维参量(均值、方差和协方差),因此,广义平稳与狭义平稳等价。
- 当高斯过程的随机变量不相关时,则它们必相互独立,故不相关与相互独立等价。
白噪声过程
S n ( f ) = N 0 / 2 S_{n}(f)=N_{0} / 2 Sn(f)=N0/2
R n ( τ ) = N 0 2 δ ( τ ) R_{n}(\tau)=\frac{N_{0}}{2} \delta(\tau) Rn(τ)=2N0δ(τ)
- 白噪声物理上不存在,因为 P = ∫ − ∞ ∞ S n ( f ) d f = ∞ P=\int_{-\infty}^{\infty}S_n(f)\mathrm{d}f=\infty P=∫−∞∞Sn(f)df=∞,但实际工程中,当噪声通过频带有限的系统,噪声功率谱带宽远大于系统带宽时,可将其近似为白噪声
- 白噪声在任何两个不同时刻的随机变量都是不相关的,具有极大的随机性
输出噪声过程 n 0 ( t ) n_0(t) n0(t)的相关函数
R n 0 ( τ ) = N 0 2 ∫ − ∞ ∞ h ( u ) h ( u + τ ) d u = N 0 2 ∫ − ∞ ∞ h ( v ) h ( v − τ ) d v \begin{aligned} R_{n_{0}}(\tau) &=\frac{N_{0}}{2} \int_{-\infty}^{\infty} h(u) h(u+\tau) \mathrm{d} u =\frac{N_{0}}{2} \int_{-\infty}^{\infty} h(v) h(v-\tau) \mathrm{d} v \end{aligned} Rn0(τ)=2N0∫−∞∞h(u)h(u+τ)du=2N0∫−∞∞h(v)h(v−τ)dv
输出噪声功率谱
S n 0 ( f ) = N 0 2 ∣ H ( f ) ∣ 2 S_{n_{0}}(f)=\frac{N_{0}}{2}|H(f)|^{2} Sn0(f)=2N0∣H(f)∣2
输出噪声功率
R n o ( 0 ) = E { n o 2 ( t ) } = N 0 2 ∫ − ∞ ∞ h 2 ( v ) d v = N 0 2 ∫ − ∞ ∞ ∣ H ( f ) ∣ 2 d f \begin{aligned} R_{n_{\mathrm{o}}}(0) &=E\left\{n_{o}^{2}(t)\right\}=\frac{N_{0}}{2} \int_{-\infty}^{\infty} h^{2}(v) \mathrm{d} v =\frac{N_{0}}{2} \int_{-\infty}^{\infty}|H(f)|^{2} \mathrm{~d} f \end{aligned} Rno(0)=E{no2(t)}=2N0∫−∞∞h2(v)dv=2N0∫−∞∞∣H(f)∣2 df
输入输出互相关函数
R n i n o ( τ ) = N 0 2 h ( τ ) R_{n_{i} n_{o}}(\tau)=\frac{N_{0}}{2} h(\tau) Rnino(τ)=2N0h(τ)
在白噪声输入时,可通过相关器来测量输入输出互相关函数,间接得到系统的冲激响应