离散傅里叶变换(Discrete Fourier Transform)
接着 离散时间傅里叶变换 讲:
-
如果 x [ n ] x[n] x[n] 只在 n = { 0 , 1 , ⋯ , N − 1 } n = \{ 0, 1, \cdots, N-1 \} n={0,1,⋯,N−1} 不为零(即有限长序列),那么
X ( ω ) = ∑ n = 0 N − 1 x [ n ] ⋅ e − j ω n X(\omega) = \sum_{n=0}^{N-1} x[n] \cdot e^{-{\rm j}\omega n} X(ω)=n=0∑N−1x[n]⋅e−jωn -
取 ω = 2 π k N , k = { 0 , 1 , ⋯ , N − 1 } \omega = \frac{2 \pi k}{N}, k = \{0, 1, \cdots, N-1 \} ω=N2πk,k={0,1,⋯,N−1},有:
X [ k ] ≜ X ( 2 π k N ) = ∑ n = 0 N − 1 x [ n ] ⋅ e − j 2 π N k n X[k] \triangleq X(\frac{2 \pi k}{N}) = \sum_{n=0}^{N-1} x[n] \cdot e^{-{\rm j}\frac{2\pi}{N}kn} X[k]≜X(N2πk)=n=0∑N−1x[n]⋅e−jN2πkn -
将上式转成矩阵相乘的形式:
x ≜ [ x [ 0 ] , x [ 1 ] , ⋯ , x [ N − 1 ] ] ⊤ \mathbf{x} \triangleq \left[ x[0], x[1], \cdots, x[N-1] \right]^\top x≜[x[0],x[1],⋯,x[N−1]]⊤
X ≜ [ X [ 0 ] , X [ 1 ] , ⋯ , X [ N − 1 ] ] ⊤ \mathbf{X} \triangleq \left[ X[0], X[1], \cdots, X[N-1] \right]^\top X≜[X[0],X[1],⋯,X[N−1]]⊤
F ≜ [ e − j 2 π N 0 ⋅ 0 e − j 2 π N 0 ⋅ 1 ⋯ e − j 2 π N 0 ⋅ ( N − 1 ) e − j 2 π N 1 ⋅ 0 e − j 2 π N 1 ⋅ 1 ⋯ e − j 2 π N 1 ⋅ ( N − 1 ) ⋮ ⋮ ⋱ ⋮ e − j 2 π N ( N − 1 ) ⋅ 0 e − j 2 π N ( N − 1 ) ⋅ 1 ⋯ e − j 2 π N ( N − 1 ) ⋅ ( N − 1 ) ] \mathbf{F} \triangleq \begin{bmatrix} e^{-{\rm j}\frac{2\pi}{N}0 \cdot 0} &e^{-{\rm j}\frac{2\pi}{N}0 \cdot 1} &\cdots &e^{-{\rm j}\frac{2\pi}{N}0 \cdot (N-1)} \\ e^{-{\rm j}\frac{2\pi}{N}1 \cdot 0} &e^{-{\rm j}\frac{2\pi}{N}1 \cdot 1} &\cdots &e^{-{\rm j}\frac{2\pi}{N}1 \cdot (N-1)}\\ \vdots &\vdots &\ddots &\vdots \\ e^{-{\rm j}\frac{2\pi}{N}(N-1) \cdot 0} &e^{-{\rm j}\frac{2\pi}{N}(N-1) \cdot 1} &\cdots &e^{-{\rm j}\frac{2\pi}{N}(N-1) \cdot (N-1)} \end{bmatrix} F≜ e−jN2π0⋅0e−jN2π1⋅0⋮e−jN2π(N−1)⋅0e−jN2π0⋅1e−jN2π1⋅1⋮e−jN2π(N−1)⋅1⋯⋯⋱⋯e−jN2π0⋅(N−1)e−jN2π1⋅(N−1)⋮e−jN2π(N−1)⋅(N−1)
X = F x \mathbf{X} = \mathbf{F} \mathbf{x} X=Fx
-
F
\mathbf{F}
F 的逆矩阵是
1
N
⋅
F
H
\frac{1}{N} \cdot \mathbf{F}^H
N1⋅FH,于是有:
x = 1 N ⋅ F H X \mathbf{x} = \frac{1}{N} \cdot \mathbf{F}^H\mathbf{X} x=N1⋅FHX
综上所述,
F
\mathbf{F}
F 是离散傅里叶变换矩阵,于是离散傅里叶变换对:
X
=
F
x
x
=
1
N
⋅
F
H
X
\mathbf{X} = \mathbf{F} \mathbf{x} \\ \mathbf{x} = \frac{1}{N} \cdot \mathbf{F}^H\mathbf{X}
X=Fxx=N1⋅FHX
或者
X
[
k
]
=
∑
n
=
0
N
−
1
x
[
n
]
⋅
e
−
j
2
π
N
k
n
x
[
n
]
=
1
N
⋅
∑
k
=
0
N
−
1
X
[
k
]
⋅
e
j
2
π
N
k
n
X[k] = \sum_{n=0}^{N-1} x[n] \cdot e^{-{\rm j}\frac{2\pi}{N}kn} \\ x[n] = \frac{1}{N} \cdot \sum_{k=0}^{N-1} X[k] \cdot e^{{\rm j}\frac{2\pi}{N}kn}
X[k]=n=0∑N−1x[n]⋅e−jN2πknx[n]=N1⋅k=0∑N−1X[k]⋅ejN2πkn
性质
有限长序列的循环变换操作
有限长序列
g
[
n
]
,
n
=
{
0
,
1
,
⋯
,
N
−
1
}
g[n], n = \{0, 1, \cdots, N-1\}
g[n],n={0,1,⋯,N−1},对其做周期为
N
N
N 的周期延拓记作
g
N
[
n
]
g_N[n]
gN[n]:
g
N
[
n
+
m
⋅
N
]
=
g
[
n
]
,
n
=
{
0
,
1
,
⋯
,
N
−
1
}
,
m
∈
Z
g_N[n+m \cdot N] = g[n], n = \{0, 1, \cdots, N-1\}, m \in \mathbb{Z}
gN[n+m⋅N]=g[n],n={0,1,⋯,N−1},m∈Z
其中,
Z
\mathbb{Z}
Z 是整数集。
同理,记另一个有限长序列
h
[
n
]
,
n
=
{
0
,
1
,
⋯
,
N
−
1
}
h[n], n = \{0, 1, \cdots, N-1\}
h[n],n={0,1,⋯,N−1},以及其周期延拓
h
N
[
n
]
h_N[n]
hN[n]。
- 循环反转
g [ n ] g[n] g[n] 经过循环反转得到 C R ( g ) [ n ] \mathbf{CR}(g)[n] CR(g)[n],具体操作过程如下:
- g ~ N [ n ] = g N [ − n ] , n ∈ Z \widetilde{g}_N[n] = g_N[-n], n \in \mathbb{Z} g N[n]=gN[−n],n∈Z
- C R ( g ) [ n ] = g ~ N [ n ] , n = { 0 , 1 , ⋯ , N − 1 } \mathbf{CR}(g)[n] = \widetilde{g}_N[n] , n = \{0, 1, \cdots, N-1\} CR(g)[n]=g N[n],n={0,1,⋯,N−1}
- 循环移位
g [ n ] g[n] g[n] 经过右循环移位 l l l 得到 C S ( g , − l ) [ n ] \mathbf{CS}(g,-l)[n] CS(g,−l)[n],具体操作过程如下:
- g ~ N [ n ] = g N [ n − l ] , n ∈ Z \widetilde{g}_N[n] = g_N[n - l], n \in \mathbb{Z} g N[n]=gN[n−l],n∈Z
- C S ( g , − l ) [ n ] = g ~ N [ n ] , n = { 0 , 1 , ⋯ , N − 1 } \mathbf{CS}(g,-l)[n] = \widetilde{g}_N[n] , n = \{0, 1, \cdots, N-1\} CS(g,−l)[n]=g N[n],n={0,1,⋯,N−1}
- 循环卷积
g [ n ] g[n] g[n] 与 h [ n ] h[n] h[n] 的循环卷积定义为:
g [ n ] ◯ h [ n ] ≜ ∑ i = 0 N − 1 g N [ i ] ⋅ h N [ n − i ] , n = { 0 , 1 , ⋯ , N − 1 } g[n] \bigcirc h[n] \triangleq \sum_{i = 0}^{N-1} g_N[i] \cdot h_N[n-i], n = \{0, 1, \cdots, N-1\} g[n]◯h[n]≜i=0∑N−1gN[i]⋅hN[n−i],n={0,1,⋯,N−1} - 循环相关
g [ n ] g[n] g[n] 与 h [ n ] h[n] h[n] 的循环相关定义为:
corr [ n ] ≜ ∑ i = 0 N − 1 g N [ i ] ⋅ h N ∗ [ i − n ] , n = { 0 , 1 , ⋯ , N − 1 } \text{corr}[n] \triangleq \sum_{i = 0}^{N-1} g_N[i] \cdot h_N^*[i-n], n = \{0, 1, \cdots, N-1\} corr[n]≜i=0∑N−1gN[i]⋅hN∗[i−n],n={0,1,⋯,N−1}
使用循环移位去实现循环相关:
- h ~ [ n ] ≜ C R ( h ) ∗ [ n ] \widetilde{h}[n] \triangleq \mathbf{CR}(h)^*[n] h [n]≜CR(h)∗[n]
- corr [ n ] = g [ n ] ◯ h ~ [ n ] \text{corr}[n] = g[n] \bigcirc \widetilde{h}[n] corr[n]=g[n]◯h [n]
性质
函数 | 离散傅里叶变换 | comments |
---|---|---|
g [ n ] g[n] g[n] | G [ k ] G[k] G[k] | |
h [ n ] h[n] h[n] | H [ k ] H[k] H[k] | |
g [ n ] ◯ h [ n ] g[n] \bigcirc h[n] g[n]◯h[n] | G [ k ] ⋅ H [ k ] G[k] \cdot H[k] G[k]⋅H[k] | 时域循环卷积 |
g [ n ] ⋅ h [ n ] g[n] \cdot h[n] g[n]⋅h[n] | 1 N ⋅ G [ k ] ◯ H [ k ] \frac{1}{N} \cdot G[k] \bigcirc H[k] N1⋅G[k]◯H[k] | 频域循环卷积 |
C S ( g , − l ) [ n ] \mathbf{CS}(g,-l)[n] CS(g,−l)[n] | e − j 2 π N k l ⋅ G [ k ] e^{-{\rm j} \frac{2 \pi}{N} k l} \cdot G[k] e−jN2πkl⋅G[k] | 时域循环移位 |
e j 2 π N k l ⋅ G [ k ] e^{{\rm j} \frac{2 \pi}{N} k l} \cdot G[k] ejN2πkl⋅G[k] | C S ( G , − l ) [ k ] \mathbf{CS}(G,-l)[k] CS(G,−l)[k] | 频域循环移位 |
g ∗ [ n ] g^*[n] g∗[n] | C R ( G ) ∗ [ k ] \mathbf{CR}(G)^*[k] CR(G)∗[k] | 时域共轭对称性 |
C R ( g ) ∗ [ n ] \mathbf{CR}(g)^*[n] CR(g)∗[n] | G ∗ [ k ] G^*[k] G∗[k] | 频域共轭对称性 |
总结
comments | |
---|---|
g ( t ) g(t) g(t) | 被采样的函数,time-limited, [ 0 , T ) [0, T) [0,T) |
G ( f ) G(f) G(f) | frequency-limited, [ − B , B ] [-B, B] [−B,B] |
T s T_s Ts | 采样周期, 1 T s ≥ 2 B \frac{1}{T_s} \geq 2B Ts1≥2B |
x [ n ] x[n] x[n] | 采样点 g ( n T s ) g(nT_s) g(nTs), n = { 0 , 1 , ⋯ , N − 1 } n = \{ 0, 1, \cdots, N-1 \} n={0,1,⋯,N−1} |
X [ k ] X[k] X[k] | H ( f ) = 1 T s ⋅ ∑ n = − ∞ + ∞ G ( f − k T s ) \displaystyle H(f) = \frac{1}{T_s} \cdot \sum_{n=-\infty}^{+\infty} G(f - \frac{k}{T_s}) H(f)=Ts1⋅n=−∞∑+∞G(f−Tsk),采样点 H ( k N T s ) , k = { 0 , 1 , ⋯ , N − 1 } H(\frac{k}{NT_s}), k = \{ 0, 1, \cdots, N-1 \} H(NTsk),k={0,1,⋯,N−1} |
x [ n ] x[n] x[n] 后面补零,提高频域分辨率 | 时域:分辨率不变,宽度增加;频域:分辨率增加,宽度不变 |
X [ k ] X[k] X[k] 中间补零,提高时域分辨率 | 时域:分辨率增加,宽度不变;频域:分辨率不变,宽度增加 |