连续信号通常是数学领域里的理论研究对象,而现实生活中我们遇到的信号往往是离散的,且计算机只能处理有限长度的离散信号。所以为了让傅立叶分析解决实际问题,有必要将其推广到离散信号领域。【阅读本文前,建议先了解连续周期信号傅里叶级数】
连续周期信号和离散周期信号如上图:左图为连续周期正弦波
x
(
t
)
=
x
(
t
+
T
)
x(t)=x(t+T)
x(t)=x(t+T),其中周期
T
=
2
π
T=2\pi
T=2π;右图为左图正弦波的周期离散采样,
x
[
n
]
=
x
[
n
+
N
]
x[n]=x[n+N]
x[n]=x[n+N],其中周期
N
=
10
N=10
N=10。
和连续周期信号傅立叶级数基于一样的猜想,离散周期信号傅立叶级数是想寻得一组不同振幅、不同频率和不同相位的正弦离散函数以表达某离散周期信号。即:
x
[
n
]
=
C
+
∑
k
=
1
∞
a
k
sin
(
k
w
0
n
)
+
∑
k
=
1
∞
b
k
cos
(
k
w
0
n
)
x[n] = C + \sum_{k=1}^{\infty}a_{k}\sin(kw_{0}n) + \sum_{k=1}^{\infty}b_{k}\cos(kw_{0}n)
x[n]=C+k=1∑∞aksin(kw0n)+k=1∑∞bkcos(kw0n)
其中
w
0
=
2
π
N
w_{0}=\frac{2\pi}{N}
w0=N2π。根据欧拉公式
e
i
x
=
cos
(
x
)
+
i
sin
(
x
)
e^{ix}=\cos(x)+i\sin(x)
eix=cos(x)+isin(x)得:
{
cos
(
x
)
=
e
i
x
+
e
−
i
x
2
sin
(
x
)
=
e
i
x
−
e
−
i
x
2
i
\begin{cases} \cos(x) = \frac{e^{ix}+e^{-ix}}{2} \\ \sin(x) = \frac{e^{ix}-e^{-ix}}{2i} \end{cases}
{cos(x)=2eix+e−ixsin(x)=2ieix−e−ix
因此,上式可推导为:
x
[
n
]
=
C
+
∑
k
=
1
∞
(
a
k
e
i
k
w
0
n
−
e
−
i
k
w
0
n
2
i
+
b
k
e
i
k
w
0
n
+
e
−
i
k
w
0
n
2
)
=
C
+
∑
k
=
1
∞
(
i
a
k
−
b
k
2
e
i
k
w
0
n
+
−
i
a
k
−
b
k
2
e
−
i
k
w
0
n
)
\begin{aligned} & x[n] = C + \sum_{k=1}^{\infty}(a_{k}\frac{e^{ikw_{0}n}-e^{-ikw_{0}n}}{2i} + b_{k}\frac{e^{ikw_{0}n}+e^{-ikw_{0}n}}{2}) \\ & = C + \sum_{k=1}^{\infty}(\frac{ia_{k}-b_{k}}{2}e^{ikw_{0}n} + \frac{-ia_{k}-b_{k}}{2}e^{-ikw_{0}n}) \\ \end{aligned}
x[n]=C+k=1∑∞(ak2ieikw0n−e−ikw0n+bk2eikw0n+e−ikw0n)=C+k=1∑∞(2iak−bkeikw0n+2−iak−bke−ikw0n)
令
A
k
=
i
a
k
−
b
k
2
A_{k} = \frac{ia_{k}-b_{k}}{2}
Ak=2iak−bk和
B
k
=
−
i
a
k
−
b
k
2
B_{k} = \frac{-ia_{k}-b_{k}}{2}
Bk=2−iak−bk,得到
x
[
n
]
的
傅
立
叶
级
数
复
数
形
式
的
表
达
式
x[n]的傅立叶级数复数形式的表达式
x[n]的傅立叶级数复数形式的表达式:
x
[
n
]
=
C
e
i
0
w
0
n
+
∑
k
=
1
∞
A
k
e
i
k
w
0
n
+
∑
k
=
1
∞
B
k
e
−
i
k
w
0
n
=
∑
k
=
−
1
−
∞
B
−
k
e
i
k
w
0
n
+
C
e
i
0
w
0
n
+
∑
k
=
1
∞
A
k
e
i
k
w
0
n
=
∑
k
=
−
∞
∞
D
k
e
i
k
w
0
n
\begin{aligned} & x[n] = Ce^{i0w_{0}n} + \sum_{k=1}^{\infty}A_{k}e^{ikw_{0}n} + \sum_{k=1}^{\infty}B_{k}e^{-ikw_{0}n} \\ & = \sum_{k=-1}^{-\infty}B_{-k}e^{ikw_{0}n} + Ce^{i0w_{0}n} + \sum_{k=1}^{\infty}A_{k}e^{ikw_{0}n} \\ & = \sum_{k=-\infty}^{\infty}D_{k}e^{ikw_{0}n} \end{aligned}
x[n]=Cei0w0n+k=1∑∞Akeikw0n+k=1∑∞Bke−ikw0n=k=−1∑−∞B−keikw0n+Cei0w0n+k=1∑∞Akeikw0n=k=−∞∑∞Dkeikw0n
我们接着观察该级数中的单项
e
i
k
w
0
n
e^{ikw_{0}n}
eikw0n:
ϕ
k
[
n
]
=
e
i
k
w
0
n
=
e
i
k
2
π
N
n
,
n
=
0
,
±
1
,
±
2
,
…
\phi_{k}[n] = e^{ikw_{0}n} = e^{ik\frac{2\pi}{N}n}, n=0,\pm1,\pm2, \dots
ϕk[n]=eikw0n=eikN2πn,n=0,±1,±2,…
先说结论:
ϕ
k
[
n
]
=
ϕ
k
+
r
N
[
n
]
\phi_{k}[n]=\phi_{k+rN}[n]
ϕk[n]=ϕk+rN[n],其中
k
=
0
,
±
1
,
±
2
,
…
k=0,\pm1,\pm2, \dots
k=0,±1,±2,…、
r
=
0
,
1
,
2
,
…
r=0,1,2,\dots
r=0,1,2,…、
N
N
N为离散信号
x
[
n
]
x[n]
x[n]的变化周期。证明过程如下:
ϕ
k
+
r
N
[
n
]
=
e
i
(
k
+
r
N
)
2
π
N
n
=
e
i
k
2
π
N
n
e
i
r
2
π
n
=
e
i
k
2
π
N
n
(
e
i
2
π
)
r
n
=
e
i
k
2
π
N
n
(
cos
(
2
π
)
+
i
sin
(
2
π
)
)
r
n
=
e
i
k
2
π
N
n
(
1
)
r
n
=
e
i
k
2
π
N
n
=
ϕ
k
[
n
]
\begin{aligned} & \phi_{k+rN}[n] = e^{i(k+rN)\frac{2\pi}{N}n}=e^{ik\frac{2\pi}{N}n}e^{ir{2\pi}n} = e^{ik\frac{2\pi}{N}n}(e^{i{2\pi}})^{rn} \\ & = e^{ik\frac{2\pi}{N}n}(\cos(2\pi) +i\sin(2\pi))^{rn} \\ & = e^{ik\frac{2\pi}{N}n}(1)^{rn} \\ & = e^{ik\frac{2\pi}{N}n} = \phi_{k}[n] \end{aligned}
ϕk+rN[n]=ei(k+rN)N2πn=eikN2πneir2πn=eikN2πn(ei2π)rn=eikN2πn(cos(2π)+isin(2π))rn=eikN2πn(1)rn=eikN2πn=ϕk[n]
因此,得到离散周期信号
x
[
n
]
x[n]
x[n]的傅立叶级数如下:
x
[
n
]
=
∑
k
=
<
N
>
X
k
e
i
k
2
π
N
n
x[n] = \sum_{k=<N>}^{}X_{k}e^{ik\frac{2\pi}{N}n}
x[n]=k=<N>∑XkeikN2πn
给出: ∑ n = < N > e i k 2 π N n = { N , k = 0 , ± N , ± 2 N , … 0 , o t h e r w i s e \sum_{n=<N>}^{}e^{ik\frac{2\pi}{N}n} = \begin{cases} N, & k=0, \pm{N}, \pm{2N}, \dots \\ 0, & otherwise \end{cases} n=<N>∑eikN2πn={N,0,k=0,±N,±2N,…otherwise 令 S = ∑ n = < N > e i k 2 π N n S = \sum_{n=<N>}^{}e^{ik\frac{2\pi}{N}n} S=∑n=<N>eikN2πn,证明过程如下: S ( 1 k N − 1 ) = e i k 2 π N S − S = e i k 2 π N N − e i k 2 π N 0 = e i k 2 π − 1 = 1 k − 1 = 0 S(1^{\frac{k}{N}}-1) = e^{ik\frac{2\pi}{N}}S-S = e^{ik\frac{2\pi}{N}N} - e^{ik\frac{2\pi}{N}0} = e^{ik{2\pi}} - 1=1^{k}-1 = 0 S(1Nk−1)=eikN2πS−S=eikN2πN−eikN2π0=eik2π−1=1k−1=0 由上式可知,当 k ≠ 0 , ± N , ± 2 N , … k \neq 0, \pm{N}, \pm{2N}, \dots k̸=0,±N,±2N,…时: ( 1 k N − 1 ) ≠ 0 且 S = ∑ n = < N > e i k 2 π N n = 0 (1^{\frac{k}{N}}-1) \neq 0 且 S = \sum_{n=<N>}^{}e^{ik\frac{2\pi}{N}n} = 0 (1Nk−1)̸=0且S=n=<N>∑eikN2πn=0 当 k = 0 , ± N , ± 2 N , … k = 0, \pm{N}, \pm{2N}, \dots k=0,±N,±2N,…时: S = ∑ n = < N > e i k 2 π N n = ∑ n = < N > ( e i 2 π ) k N n = ∑ n = < N > ( 1 ) r n = N , r = 0 , ± 1 , ± 2 , … S = \sum_{n=<N>}^{}e^{ik\frac{2\pi}{N}n} = \sum_{n=<N>}^{}(e^{i2\pi} )^{\frac{k}{N}n} = \sum_{n=<N>}^{}(1)^{rn} = N, r = 0,\pm1,\pm2,\dots S=n=<N>∑eikN2πn=n=<N>∑(ei2π)Nkn=n=<N>∑(1)rn=N,r=0,±1,±2,…
现在,对离散周期信号
x
[
n
]
x[n]
x[n]每时刻的信号求和,并乘以
e
−
i
r
2
π
N
n
e^{-ir\frac{2\pi}{N}n}
e−irN2πn得:
∑
n
=
<
N
>
x
[
n
]
e
−
i
r
2
π
N
n
=
∑
n
=
<
N
>
∑
k
=
<
N
>
X
k
e
i
(
k
−
r
)
2
π
N
n
=
∑
k
=
<
N
>
X
k
∑
n
=
<
N
>
e
i
(
k
−
r
)
2
π
N
n
\sum_{n=<N>}^{} x[n] e^{-ir\frac{2\pi}{N}n} = \sum_{n=<N>}^{} \sum_{k=<N>}^{}X_{k}e^{i(k-r)\frac{2\pi}{N}n} = \sum_{k=<N>}^{} X_{k} \sum_{n=<N>}^{}e^{i(k-r)\frac{2\pi}{N}n}
n=<N>∑x[n]e−irN2πn=n=<N>∑k=<N>∑Xkei(k−r)N2πn=k=<N>∑Xkn=<N>∑ei(k−r)N2πn
由上面给出的公式可以得出,当
k
−
r
=
0
k-r=0
k−r=0时,
∑
n
=
<
N
>
e
i
(
k
−
r
)
2
π
N
n
=
N
\sum_{n=<N>}^{}e^{i(k-r)\frac{2\pi}{N}n} = N
∑n=<N>ei(k−r)N2πn=N;当
k
−
r
≠
0
k-r \neq 0
k−r̸=0时,
∑
n
=
<
N
>
e
i
(
k
−
r
)
2
π
N
n
=
0
\sum_{n=<N>}^{}e^{i(k-r)\frac{2\pi}{N}n} = 0
∑n=<N>ei(k−r)N2πn=0。所以:
∑
n
=
<
N
>
x
[
n
]
e
−
i
r
2
π
N
n
=
∑
k
=
<
N
>
X
k
∑
n
=
<
N
>
e
i
(
k
−
r
)
2
π
N
n
=
X
r
N
\sum_{n=<N>}^{} x[n] e^{-ir\frac{2\pi}{N}n} = \sum_{k=<N>}^{} X_{k} \sum_{n=<N>}^{}e^{i(k-r)\frac{2\pi}{N}n} = X_{r} N
n=<N>∑x[n]e−irN2πn=k=<N>∑Xkn=<N>∑ei(k−r)N2πn=XrN
即:
X
r
=
1
N
∑
n
=
<
N
>
x
[
n
]
e
−
i
r
2
π
N
n
X_{r} = \frac{1}{N} \sum_{n=<N>}^{} x[n] e^{-ir\frac{2\pi}{N}n}
Xr=N1n=<N>∑x[n]e−irN2πn
至此,我们已经得到离散周期信号
x
[
n
]
x[n]
x[n]的傅立叶级数(如下):
x
[
n
]
=
∑
k
=
<
N
>
X
k
e
i
k
2
π
N
n
x[n] = \sum_{k=<N>}^{}X_{k}e^{ik\frac{2\pi}{N}n}
x[n]=k=<N>∑XkeikN2πn
其中, X k = 1 N ∑ n = < N > x [ n ] e − i k 2 π N n X_{k} = \frac{1}{N} \sum_{n=<N>}^{} x[n] e^{-ik\frac{2\pi}{N}n} Xk=N1∑n=<N>x[n]e−ikN2πn。