Paseval定理与功率谱密度
Paseval定理:
假定A(x),B(x)以2π为周期,且有变换:$A(x) = \sum\limits_{n = - \infty }^\infty {{a_n}{e^{inx}}} $,$B(x) = \sum\limits_{n = - \infty }^\infty {{b_n}{e^{inx}}}$
则有等式:
\[\sum\limits_{n = - \infty }^\infty {{a_n}{{\bar b}_n}} = \frac{1}{{2\pi }}\int_{ - \pi }^\pi {A(x)\overline {B(x)} dx} \]
上述定理是对傅里叶级数来说的,对于傅里叶变换,则有Plancherel 定理:
\[\int_{ - \infty }^\infty {f(x)\overline {g(x)} dx = } \int_{ - \infty }^\infty {\hat f(\xi )\overline {\hat g(\xi )} d\xi } \]
其中
\[\hat f(\xi ) = \int_{ - \infty }^\infty {f(x){e^{ - 2\pi i\xi x}}dx} \]
由上述两个定理,对四种变换可得到4个等式:
FT :\[\int_{ - \infty }^\infty {f{{(t)}^2}dt} = \frac{1}{{2\pi }}\int_{ - \infty }^\infty {F{{(\omega )}^2}d\omega = } \int_{ - \infty }^\infty {F{{(2\pi f)}^2}df} \]
DTFT:\[\sum\limits_{n = - \infty }^\infty {x_n^2 = \frac{1}{{2\pi }}} \int_{ - \pi }^\pi {F(\omega )^2d\omega } \]
FS:\[\frac{1}{T}\int_{ - T/2}^{T/2} {{f^2}(t)dt = \sum\limits_{k = - \infty }^\infty {C_k^2} } \]
DFS:\[\sum\limits_{n = 0}^{N - 1} {x_n^2 = \frac{1}{N}} \sum\limits_{k = 0}^{N - 1} {X_k^2} \]
这即是常见的时域频域等式,也常常被称为Paseval定理。
对于FS变换,Paseval等式也可从FT中导出,截取一个周期做FT变换,并等式两边同除以T可得:
\[\frac{1}{T}\int_{ - T/2}^{T/2} {f{{(t)}^2}dt} = \frac{1}{{2\pi }}\int_{ - \infty }^\infty {\frac{{F{{(\omega )}^2}}}{T}d\omega } \]
其中,\[F(\omega ) = \int_{ - T/2}^{T/2} {f(t){e^{ - i\omega t}}dt} \]
对比:\[{C_k} = \frac{1}{T}\int_{ - T/2}^{T/2} {f(t){e^{ - i{\omega _k}t}}dt} \]
则:Ck=F(ωk)/T
从而:
\[\begin{gathered}
\frac{1}{{2\pi }}\int_{ - \infty }^\infty {\frac{{F{{(\omega )}^2}}}{T}d\omega } = \sum\limits_{k = - \infty }^\infty {\int_{k/T}^{(k + 1)/T} {\frac{{F{{(\omega )}^2}}}{T}df} } = \sum\limits_{k = - \infty }^\infty {\frac{{F{{({\omega _k})}^2}}}{T}\Delta f} \hfill \\
= \sum\limits_{k = - \infty }^\infty {\frac{{F{{({\omega _k})}^2}}}{T}\Delta f} = \sum\limits_{k = - \infty }^\infty {C_k^2} \hfill \\
\end{gathered} \]
而DFS的等式形式也可以由FS等式推导,令dt=T/N,Ck=Xk/N就可得到。
对于功率信号,注意到FT要求$\int_{ - \infty }^\infty {\left| {f(t)} \right|dt < \infty } $,而功率信号并不满足这个条件,但考虑到功率信号的功率有限,则有等式:
\[P = \mathop {\lim }\limits_{T \to \infty } \frac{1}{T}\int_{ - \frac{T}{2}}^{\frac{T}{2}} f {(t)^2}dt = \mathop {\lim }\limits_{T \to \infty } \frac{1}{{2\pi }}\int {\frac{{F{{(\omega )}^2}}}{T}} d\omega \]
这即是功率信号所满足的Paseval定理。
单位频率上的能量称为能量谱密度,单位频率上的功率称为功率谱密度,由上述Paseval等式,可以定义:
能量谱密度:\[E(\omega ) = F{(\omega )^2} = F(\omega ){F^*}(\omega )\]
功率谱密度:\[P(\omega ) = \mathop {\lim }\limits_{T \to \infty } \frac{{F{{(\omega )}^2}}}{T} = \mathop {\lim }\limits_{T \to \infty } \frac{{F(\omega ){F^*}(\omega )}}{T}\]
由卷积的性质:$FT({f_1}(t) * {f_2}(t)) = {F_1}(\omega ){F_2}(\omega )$和f(-t)的FT变换为F*(ω),则可得到
\[F(\omega ){F^*}(\omega ) = FT(f(t) * f( - t))\]
对于能量信号,假如定义自相关函数\[R(\tau ) = \int_{ - \infty }^\infty {x(t)x(t + \tau )dt = x(t) * x( - t)} \]
而对于功率信号,假如定义自相关函数\[R(\tau ) = E\left[ {\int_{ - \infty }^\infty {x(t)x(t + \tau )dt} } \right] = E\left[ {x(t) * x( - t)} \right]\]
则能量谱密度或功率谱密度即为自相关函数R(τ)的FT变换,且R(0)即为信号的能量或功率。
对于DFT变换,与DFS一样,满足:
\[\sum\limits_{n = 0}^{N - 1} {x_n^2 = \frac{1}{N}} \sum\limits_{k = 0}^{N - 1} {X_k^2} \]
等式两边同除以N,有
\[\frac{1}{N}\sum\limits_{n} x_n^2=\frac{1}{N^2}\sum\limits_{k} X_k^2\]
等式左边为信号的功率,所以右边也为信号功率,$\frac{X_k^2}{N^2}$则可以视作频率$\frac{k}{N} f_s$处,带宽fs/N内信号平均功率,这样功率谱密度为:
\[P(\omega ) = \frac{{X_k^2/{N^2}}}{{{f_s}/N}} = \frac{{X_k^2}}{{N{f_s}}}\]
例子:
假定有一数字序列
\( x_n=cos(\omega_1 n)+cos(\omega_2 n)+N(0,0.01)\)
其中,ω1=2π*1/50,ω2=2π*5/50,N为均值为0,方差为0.01的高斯白噪声
信号的功率:Power(x)=var(x)=1.0074(watt)
数字频率为ω=0~2π,所以,等价的单边带噪声平均功率谱密度Px=Power/π=0.3206(watt/rad)
1)用xn的傅里叶变换来计算单边带功率谱密度,公式为$P_{\omega}=\frac{X_K^2}{Nf_s/2}=\frac{X_K^2}{N\pi}$
结果为
平均功率谱密度为:0.3265(watt/rad)
2)用Welch方法估计为:
平均功率谱密度为:0.3191(watt/rad)
3)自相关方法计算功率谱密度:
1. N=length(x); 2. cx=xcorr(x,'biased'); 3. zcx=fft(cx(N:end)); 4. zcx=real(zcx)/pi;%%利用cx作为偶函数的对称性 5. plot(linspace(0,pi,N),10*log10(zcx))
平均功率谱密度为:0.3206(watt/rad)