Josh 的复习总结之数字信号处理(Part 2——离散傅里叶级数 DFS)

闲来无事编点程 同时被 2 个专栏收录
7 篇文章 1 订阅
7 篇文章 1 订阅


《Josh 的复习总结之数字信号处理》系列文章目录:

      Part 1——离散时间信号和系统分析基础
👉 Part 2——离散傅里叶级数 DFS
      Part 3——离散傅里叶变换 DFT
      Part 4——快速傅里叶变换 FFT
      Part 5——部分 FFT 蝶形图
      Part 6——数字滤波器的基本结构
      Part 7——数字滤波器设计


1. 傅里叶变换的几种形式

1.1 非周期连续时间信号的傅里叶变换

  非周期连续时间信号的傅里叶变换(连续时间傅里叶变换,Continuous-Time Fourier Transform, CTFT)的形式为:
X a ( j Ω ) = ∫ − ∞ ∞ x a ( t ) e − j Ω t d t X_a\left(j\Omega\right)=\int_{-\infty}^{\infty}{x_a\left(t\right)e^{-j\Omega t}\mathrm{d}t} Xa(jΩ)=xa(t)ejΩtdt x a ( t ) = 1 2 π ∫ − ∞ ∞ X a ( j Ω ) e j Ω t d Ω x_a\left(t\right)=\frac{1}{2\pi}\int_{-\infty}^{\infty}{X_a\left(j\Omega\right)e^{j\Omega t}\mathrm{d}\Omega} xa(t)=2π1Xa(jΩ)ejΩtdΩ

1.2 周期连续时间信号的傅里叶变换

  周期连续时间信号的傅里叶变换(连续傅里叶级数,Continous Fourier Series, CFS)的形式为:
X ( m Ω ) = 1 t p ∫ − t p 2 t p 2 x a ( t ) e − j m Ω t d t X\left(m\Omega\right)=\frac{1}{t_p}\int_{-\frac{t_p}{2}}^{\frac{t_p}{2}}{x_a\left(t\right)e^{-jm\Omega t}\mathrm{d}t} X(mΩ)=tp12tp2tpxa(t)ejmΩtdt x a ( t ) = ∑ m = − ∞ ∞ X ( m Ω ) e j m Ω t x_a\left(t\right)=\sum_{m=-\infty}^{\infty}{X\left(m\Omega\right)e^{jm\Omega t}} xa(t)=m=X(mΩ)ejmΩt其中 t p t_p tp 为信号的周期,频域相邻谱线间角频率增量 Ω = 2 π t p \Omega=\dfrac{2\pi}{t_p} Ω=tp2π

1.3 非周期离散时间信号的傅里叶变换

  非周期离散时间信号的傅里叶变换(离散时间傅里叶变换,Discrete-Time Fourier Transform, DTFT)的形式为:
X ( e j ω ) = ∑ n = − ∞ ∞ x ( n ) e − j ω n X\left(e^{j\omega}\right)=\sum_{n=-\infty}^{\infty}{x\left(n\right)e^{-j\omega n}} X(ejω)=n=x(n)ejωn x ( n ) = 1 2 π ∫ − π π X ( e j ω ) e j ω n d ω x\left(n\right)=\frac{1}{2\pi}\int_{-\pi}^{\pi}{X\left(e^{j\omega}\right)e^{j\omega n}\mathrm{d}\omega} x(n)=2π1ππX(ejω)ejωndω取样频率 f s = 1 T f_s=\dfrac{1}{T} fs=T1,取样角频率 Ω s = 2 π T \Omega_s=\dfrac{2\pi}{T} Ωs=T2π,取样数字频率 ω s = 2 π \omega_s=2\pi ωs=2π

1.4 周期离散时间信号的傅里叶变换

  周期离散时间信号的傅里叶变换(离散傅里叶级数,Discrete Fourier Series, DFS)的形式为:
X ~ ( k ) = X ~ ( e j ω ) ∣ ω = k 2 π N = ∑ n = 0 N − 1 x ~ ( n ) e − j 2 π N k n = ∑ n = 0 N − 1 x ~ ( n ) W N k n \widetilde{X}\left(k\right)=\left.\widetilde{X}\left(e^{j\omega}\right)\right|_{\omega=k\frac{2\pi}{N}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)W_N^{kn}} X (k)=X (ejω)ω=kN2π=n=0N1x (n)ejN2πkn=n=0N1x (n)WNkn x ~ ( n ) = 1 N ∑ k = 0 N − 1 X ~ ( k ) e j 2 π N k n = 1 N ∑ k = 0 N − 1 X ~ ( k ) W N − k n \widetilde{x}\left(n\right)=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kn}}=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)W_N^{-kn}} x (n)=N1k=0N1X (k)ejN2πkn=N1k=0N1X (k)WNkn其中 N N N 为一个周期内的点数,时域相邻谱线间隔为 T T T ,频域相邻谱线间隔为 1 N T \dfrac{1}{NT} NT1,以数字频率表征时为 2 π N \dfrac{2\pi}{N} N2π

由以上,在一个域中对函数进行取样,两取样点间增量的倒数,必是另一个域中的周期。


2. 离散傅里叶级数(DFS)的推导

2.1 正变换

  令 x ( n ) x\left(n\right) x(n) N N N 点长的离散序列,则其 DTFT(上标“~”用来表征周期性)
X ~ ( e j ω ) = ∑ n = − ∞ ∞ x ( n ) e − j ω n = ∑ n = 0 N − 1 x ( n ) e − j ω n \widetilde{X}\left(e^{j\omega}\right)=\sum_{n=-\infty}^{\infty}{x\left(n\right)e^{-j\omega n}}=\sum_{n=0}^{N-1}{x\left(n\right)e^{-j\omega n}} X (ejω)=n=x(n)ejωn=n=0N1x(n)ejωn X ~ ( e j ω ) \widetilde{X}\left(e^{j\omega}\right) X (ejω) 取样,使其称为周期离散频率函数,由此时域序列 x ( n ) x\left(n\right) x(n) 周期化为 x ~ ( n ) \widetilde{x}\left(n\right) x (n),令其谱线间隔为 T T T,则时域序列的周期为 N T NT NT,所以对频谱取样的谱间距为 1 N T \dfrac{1}{NT} NT1,以数字频率表征时,谱间距
ω I = 2 π N \omega_\mathrm{I}=\frac{2\pi}{N} ωI=N2π因此,以数字频率 ω \omega ω 为变量的 X ( e j ω ) X\left(e^{j\omega}\right) X(ejω) 被离散化时,其变量 ω \omega ω 变为
ω = k ω I = k 2 π N \omega=k\omega_\mathrm{I}=k\frac{2\pi}{N} ω=kωI=kN2π因此离散周期序列 x ~ ( n ) \widetilde{x}\left(n\right) x (n) 的傅里叶级数可写为
X ~ ( k ) = X ~ ( e j ω ) ∣ ω = k 2 π N = ∑ n = 0 N − 1 x ~ ( n ) e − j 2 π N k n = ∑ n = 0 N − 1 x ~ ( n ) W N k n \widetilde{X}\left(k\right)=\left.\widetilde{X}\left(e^{j\omega}\right)\right|_{\omega=k\frac{2\pi}{N}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)W_N^{kn}} X (k)=X (ejω)ω=kN2π=n=0N1x (n)ejN2πkn=n=0N1x (n)WNkn并将数字域频率用 k k k 简化表示。且
X ~ ( k + N ) = ∑ n = 0 N − 1 x ~ ( n ) e − j 2 π N ( k + N ) n = ∑ n = 0 N − 1 x ~ ( n ) e − j 2 π n e − j 2 π N k n = ∑ n = 0 N − 1 x ~ ( n ) e − j 2 π N k n = X ~ ( k ) \widetilde{X}\left(k+N\right)=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}\left(k+N\right)n}}=\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j2\pi n}e^{-j\frac{2\pi}{N}kn}=}\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}=}\widetilde{X}\left(k\right) X (k+N)=n=0N1x (n)ejN2π(k+N)n=n=0N1x (n)ej2πnejN2πkn=n=0N1x (n)ejN2πkn=X (k) X ~ ( k ) \widetilde{X}\left(k\right) X (k) 是以 N N N 为周期的周期函数。

2.2 逆变换

  将正变换两边乘以 e j 2 π N k r e^{j\frac{2\pi}{N}kr} ejN2πkr,并在一个周期内求和,有
∑ k = 0 N − 1 X ~ ( k ) e j 2 π N k r = ∑ k = 0 N − 1 ( ∑ n = 0 N − 1 x ~ ( n ) e − j 2 π N k n ) e j 2 π N k r = N [ ∑ n = 0 N − 1 x ~ ( n ) ( 1 N ∑ k = 0 N − 1 e j 2 π N k ( r − n ) ) ] \sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kr}}=\sum_{k=0}^{N-1}{\left(\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)e^{-j\frac{2\pi}{N}kn}}\right)e^{j\frac{2\pi}{N}kr}}=N\left[\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)\left(\frac{1}{N}\sum_{k=0}^{N-1}e^{j\frac{2\pi}{N}k\left(r-n\right)}\right)}\right] k=0N1X (k)ejN2πkr=k=0N1(n=0N1x (n)ejN2πkn)ejN2πkr=N[n=0N1x (n)(N1k=0N1ejN2πk(rn))]由正交定理
1 N ∑ k = 0 N − 1 e j 2 π N k ( r − n ) = { 1 , r = n 0 , r ≠ n \frac{1}{N}\sum_{k=0}^{N-1}e^{j\frac{2\pi}{N}k\left(r-n\right)}= \begin{aligned} \begin{cases} 1,&r=n\\ 0,&r≠n \end{cases} \end{aligned} N1k=0N1ejN2πk(rn)={1,0,r=nr=n
∑ k = 0 N − 1 X ~ ( k ) e j 2 π N k r = N [ ∑ n = 0 N − 1 x ~ ( n ) ] ∣ n = r = N x ~ ( r ) \sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kr}}=N\left.\left[\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)}\right]\right|_{n=r}=N\widetilde{x}\left(r\right) k=0N1X (k)ejN2πkr=N[n=0N1x (n)]n=r=Nx (r)可得逆离散傅里叶变换
x ~ ( n ) = 1 N ∑ k = 0 N − 1 X ~ ( k ) e j 2 π N k n = 1 N ∑ k = 0 N − 1 X ~ ( k ) W N − k n \widetilde{x}\left(n\right)=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kn}}=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)W_N^{-kn}} x (n)=N1k=0N1X (k)ejN2πkn=N1k=0N1X (k)WNkn
x ~ ( n + m N ) = 1 N ∑ k = 0 N − 1 X ~ ( k ) e j 2 π N k ( n + m N ) = 1 N ∑ k = 0 N − 1 X ~ ( k ) e j 2 π N k n e j 2 π k m = x ~ ( n ) \widetilde{x}\left(n+mN\right)=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}k\left(n+mN\right)}}=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)e^{j\frac{2\pi}{N}kn}e^{j2\pi km}}=\widetilde{x}\left(n\right) x (n+mN)=N1k=0N1X (k)ejN2πk(n+mN)=N1k=0N1X (k)ejN2πknej2πkm=x (n) x ~ ( n ) \widetilde{x}\left(n\right) x (n) 是以 N N N 为周期的周期函数。


3. 离散傅里叶级数(DFS)的性质

3.1 线性性质

D F S [ a x ~ 1 ( n ) + b x ~ 2 ( n ) ] = a X ~ 1 ( k ) + b X ~ 2 ( k ) \mathrm{DFS}\left[a{\widetilde{x}}_1\left(n\right)+{b\widetilde{x}}_2\left(n\right)\right]=a{\widetilde{X}}_1\left(k\right)+b{\widetilde{X}}_2\left(k\right) DFS[ax 1(n)+bx 2(n)]=aX 1(k)+bX 2(k)

3.2 时移性质

D F S [ x ~ ( n + m ) ] = W N − m k X ~ ( k ) \mathrm{DFS}\left[\widetilde{x}\left(n+m\right)\right]=W_N^{-mk}\widetilde{X}\left(k\right) DFS[x (n+m)]=WNmkX (k)证明:
D F S [ x ~ ( n + m ) ] = ∑ n = 0 N − 1 x ~ ( n + m ) W N n k = ∑ n = m N − 1 + m x ~ ( n ) W N ( n − m ) k = ∑ n = m N − 1 + m x ~ ( n ) W N n k W N − m k = 利 用 x ~ ( n ) 和 W N n k 的 周 期 性 ( ∑ n = 0 N − 1 x ~ ( n ) W N n k ) W N − m k = W N − m k X ~ ( k ) \begin{aligned} \mathrm{DFS}\left[\widetilde{x}\left(n+m\right)\right]&=\sum_{n=0}^{N-1}{\widetilde{x}\left(n+m\right)W_N^{nk}}=\sum_{n=m}^{N-1+m}{\widetilde{x}\left(n\right)W_N^{\left(n-m\right)k}}=\sum_{n=m}^{N-1+m}{\widetilde{x}\left(n\right)W_N^{nk}W_N^{-mk}} \\ &\xlongequal{利用\widetilde{x}\left(n\right)和W_N^{nk}的周期性}\left(\sum_{n=0}^{N-1}{\widetilde{x}\left(n\right)W_N^{nk}}\right)W_N^{-mk}=W_N^{-mk}\widetilde{X}\left(k\right) \end{aligned} DFS[x (n+m)]=n=0N1x (n+m)WNnk=n=mN1+mx (n)WN(nm)k=n=mN1+mx (n)WNnkWNmkx (n)WNnk (n=0N1x (n)WNnk)WNmk=WNmkX (k)

3.3 频移性质

I D F S [ X ~ ( k + l ) ] = W N n l x ~ ( n ) \mathrm{IDFS}\left[\widetilde{X}\left(k+l\right)\right]=W_N^{nl}\widetilde{x}\left(n\right) IDFS[X (k+l)]=WNnlx (n)证明:
I D F S [ X ~ ( k + l ) ] = 1 N ∑ k = 0 N − 1 X ~ ( k + l ) W N − k n = 1 N ∑ k = l N − 1 + l X ~ ( k ) W N − ( k − l ) n = 1 N ∑ k = l N − 1 + l X ~ ( k ) W N − k n W N n l = 利 用 X ~ ( n ) 和 W N − n k 的 周 期 性 1 N ( ∑ k = 0 N − 1 X ~ ( k ) W N − k n ) W N n l = W N n l x ~ ( n ) \begin{aligned} \mathrm{IDFS}\left[\widetilde{X}\left(k+l\right)\right]&=\frac{1}{N}\sum_{k=0}^{N-1}{\widetilde{X}\left(k+l\right)W_N^{-kn}}=\frac{1}{N}\sum_{k=l}^{N-1+l}{\widetilde{X}\left(k\right)W_N^{-\left(k-l\right)n}}=\frac{1}{N}\sum_{k=l}^{N-1+l}{\widetilde{X}\left(k\right)W_N^{-kn}W_N^{nl}} \\ &\xlongequal{利用\widetilde{X}\left(n\right)和W_N^{-nk}的周期性}\frac{1}{N}\left(\sum_{k=0}^{N-1}{\widetilde{X}\left(k\right)W_N^{-kn}}\right)W_N^{nl}=W_N^{nl}\widetilde{x}\left(n\right) \end{aligned} IDFS[X (k+l)]=N1k=0N1X (k+l)WNkn=N1k=lN1+lX (k)WN(kl)n=N1k=lN1+lX (k)WNknWNnlX (n)WNnk N1(k=0N1X (k)WNkn)WNnl=WNnlx (n)

3.4 时域周期卷积

x ~ 1 ( n ) ⊙ x ~ 2 ( n ) = I D F S [ X ~ 1 ( k ) ⋅ X ~ 2 ( k ) ] {\widetilde{x}}_1\left(n\right)\odot{\widetilde{x}}_2\left(n\right)=\mathrm{IDFS}\left[{\widetilde{X}}_1\left(k\right)\cdot{\widetilde{X}}_2\left(k\right)\right] x 1(n)x 2(n)=IDFS[X 1(k)X 2(k)]证明:
I D F S [ X ~ 1 ( k ) ⋅ X ~ 2 ( k ) ] = 1 N ∑ k = 0 N − 1 [ X ~ 1 ( k ) ⋅ X ~ 2 ( k ) ] W N − n k = 1 N ∑ k = 0 N − 1 [ ∑ m = 0 N − 1 x ~ 1 ( m ) W N m k ] X ~ 2 ( k ) W N − n k = 1 N ∑ k = 0 N − 1 [ ∑ m = 0 N − 1 x ~ 1 ( m ) X ~ 2 ( k ) W N − ( n − m ) k ] = ∑ m = 0 N − 1 x ~ 1 ( m ) [ 1 N ∑ k = 0 N − 1 X ~ 2 ( k ) W N − ( n − m ) k ] = ∑ m = 0 N − 1 x ~ 1 ( m ) x ~ 2 ( n − m ) = x ~ 1 ( n ) ⊙ x ~ 2 ( n ) \begin{aligned} \mathrm{IDFS}\left[{\widetilde{X}}_1\left(k\right)\cdot{\widetilde{X}}_2\left(k\right)\right]&=\frac{1}{N}\sum_{k=0}^{N-1}{\left[{\widetilde{X}}_1\left(k\right)\cdot{\widetilde{X}}_2\left(k\right)\right]W_N^{-nk}}\\ &=\frac{1}{N}\sum_{k=0}^{N-1}{\left[\sum_{m=0}^{N-1}{{\widetilde{x}}_1\left(m\right)W_N^{mk}}\right]{\widetilde{X}}_2\left(k\right)W_N^{-nk}}\\ &=\frac{1}{N}\sum_{k=0}^{N-1}\left[\sum_{m=0}^{N-1}{{\widetilde{x}}_1\left(m\right){{\widetilde{X}}_2\left(k\right)W}_N^{-\left(n-m\right)k}}\right]\\ &=\sum_{m=0}^{N-1}{{\widetilde{x}}_1\left(m\right)\left[\frac{1}{N}\sum_{k=0}^{N-1}{{\widetilde{X}}_2\left(k\right)W}_N^{-\left(n-m\right)k}\right]}\\ &=\sum_{m=0}^{N-1}{{\widetilde{x}}_1\left(m\right){\widetilde{x}}_2\left(n-m\right)} \\ &={\widetilde{x}}_1\left(n\right)\odot{\widetilde{x}}_2\left(n\right) \end{aligned} IDFS[X 1(k)X 2(k)]=N1k=0N1[X 1(k)X 2(k)]WNnk=N1k=0N1[m=0N1x 1(m)WNmk]X 2(k)WNnk=N1k=0N1[m=0N1x 1(m)X 2(k)WN(nm)k]=m=0N1x 1(m)[N1k=0N1X 2(k)WN(nm)k]=m=0N1x 1(m)x 2(nm)=x 1(n)x 2(n)

注:此处用 ⊙ \odot 表示圆周卷积,不同教材中圆周卷积的符号表述可能不同。

3.5 频域周期卷积

D F S [ x ~ 1 ( n ) x ~ 2 ( n ) ] = 1 N [ X ~ 1 ( k ) ⊙ X ~ 2 ( k ) ] \mathrm{DFS}\left[{\widetilde{x}}_1\left(n\right){\widetilde{x}}_2\left(n\right)\right]=\frac{1}{N}\left[{\widetilde{X}}_1\left(k\right)\odot{\widetilde{X}}_2\left(k\right)\right] DFS[x 1(n)x 2(n)]=N1[X 1(k)X 2(k)]证明:
D F S [ x ~ 1 ( n ) x ~ 2 ( n ) ] = ∑ n = 0 N − 1 x ~ 1 ( n ) x ~ 2 ( n ) W N k n = ∑ n = 0 N − 1 [ 1 N ∑ r = 0 N − 1 X ~ 1 ( r ) W N − r n ] x ~ 2 ( n ) W N k n = 1 N ∑ n = 0 N − 1 [ ∑ r = 0 N − 1 X ~ 1 ( r ) x ~ 2 ( n ) W N ( k − r ) n ] = 1 N ∑ r = 0 N − 1 X ~ 1 ( r ) [ ∑ k = 0 N − 1 x ~ 2 ( n ) W N ( k − r ) n ] = 1 N ∑ r = 0 N − 1 X ~ 1 ( r ) X ~ 2 ( k − r ) = 1 N [ X ~ 1 ( k ) ⊙ X ~ 2 ( k ) ] \begin{aligned} \mathrm{DFS}\left[{\widetilde{x}}_1\left(n\right){\widetilde{x}}_2\left(n\right)\right]&=\sum_{n=0}^{N-1}{{\widetilde{x}}_1\left(n\right){\widetilde{x}}_2\left(n\right)W_N^{kn}}=\sum_{n=0}^{N-1}{\left[\frac{1}{N}\sum_{r=0}^{N-1}{{\widetilde{X}}_1\left(r\right)W_N^{-rn}}\right]{\widetilde{x}}_2\left(n\right)W_N^{kn}}\\ &=\frac{1}{N}\sum_{n=0}^{N-1}\left[\sum_{r=0}^{N-1}{{\widetilde{X}}_1\left(r\right){\widetilde{x}}_2\left(n\right)W_N^{\left(k-r\right)n}}\right]\\ &=\frac{1}{N}\sum_{r=0}^{N-1}{{\widetilde{X}}_1\left(r\right)\left[\sum_{k=0}^{N-1}{{\widetilde{x}}_2\left(n\right)W_N^{\left(k-r\right)n}}\right]}\\ &=\frac{1}{N}\sum_{r=0}^{N-1}{{\widetilde{X}}_1\left(r\right){\widetilde{X}}_2\left(k-r\right)}\\ &=\frac{1}{N}\left[{\widetilde{X}}_1\left(k\right)\odot{\widetilde{X}}_2\left(k\right)\right] \end{aligned} DFS[x 1(n)x 2(n)]=n=0N1x 1(n)x 2(n)WNkn=n=0N1[N1r=0N1X 1(r)WNrn]x 2(n)WNkn=N1n=0N1[r=0N1X 1(r)x 2(n)WN(kr)n]=N1r=0N1X 1(r)[k=0N1x 2(n)WN(kr)n]=N1r=0N1X 1(r)X 2(kr)=N1[X 1(k)X 2(k)]

注:此处用 ⊙ \odot 表示圆周卷积,不同教材中圆周卷积的符号表述可能不同。


上一篇:Part 1——离散时间信号和系统分析基础
下一篇:Part 3——离散傅里叶变换 DFT

  • 1
    点赞
  • 0
    评论
  • 0
    收藏
  • 一键三连
    一键三连
  • 扫一扫,分享海报

相关推荐
©️2020 CSDN 皮肤主题: 猿与汪的秘密 设计师:白松林 返回首页
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、C币套餐、付费专栏及课程。

余额充值