引用自傅里叶系列(二)傅里叶变换的推导,在此基础上对一些细节重新整理了一下
1、先把傅里叶级数转为指数形式
三角函数形式:
f
(
t
)
=
a
0
2
+
∑
n
=
1
∞
(
a
n
c
o
s
(
n
ω
t
)
+
b
n
s
i
n
(
n
ω
t
)
)
)
a
0
=
2
T
∫
t
0
t
0
+
T
f
(
t
)
d
t
a
n
=
2
T
∫
t
0
t
0
+
T
f
(
t
)
c
o
s
(
n
ω
t
)
d
t
b
n
=
2
T
∫
t
0
t
0
+
T
f
(
t
)
s
i
n
(
n
ω
t
)
d
t
(1)
\begin{aligned} f(t) &= \frac{a_0}{2}+\sum_{n=1}^{\infty}(a_n cos(n\omega t)+b_nsin(n \omega t)) \tag{1}) \\ a_0 & = \frac{2}{T}\int_{t_0}^{t_0+T}f(t)dt \\ a_n &= \frac{2}{T}\int_{t_0}^{t_0+T}f(t)cos(n\omega t)dt \\ b_n &= \frac{2}{T}\int_{t_0}^{t_0+T}f(t)sin(n\omega t)dt \\ \end{aligned}
f(t)a0anbn=2a0+n=1∑∞(ancos(nωt)+bnsin(nωt)))=T2∫t0t0+Tf(t)dt=T2∫t0t0+Tf(t)cos(nωt)dt=T2∫t0t0+Tf(t)sin(nωt)dt(1)
上式中:
ω
=
2
π
T
\omega=\frac{2\pi}{T}
ω=T2π
式中,
T
T
T为函数
f
(
t
)
f(t)
f(t)的周期,往往是趋向于
+
∞
+\infty
+∞的,因此
ω
\omega
ω则趋向于0,因为这样的关系,当函数
f
(
t
)
f(t)
f(t)的周期越大时,傅里叶变换后的频率越密集。
欧拉公式:
e
i
θ
=
c
o
s
(
θ
)
+
i
s
i
n
(
θ
)
e
−
i
θ
=
c
o
s
(
θ
)
−
i
s
i
n
(
θ
)
(2)
\begin{aligned} e^{i\theta}=cos(\theta)+i sin(\theta) \\ e^{-i \theta}=cos(\theta)-i sin(\theta) \\ \tag{2} \end{aligned}
eiθ=cos(θ)+isin(θ)e−iθ=cos(θ)−isin(θ)(2)
因此,可得
c
o
s
(
θ
)
=
e
i
θ
+
e
−
i
θ
2
s
i
n
(
θ
)
=
−
i
⋅
e
i
θ
−
e
−
i
θ
2
(3)
\begin{aligned} cos(\theta)&=\frac{e^{i\theta} + e^{-i\theta}}{2} \\ sin(\theta)&=-i\cdot\frac{e^{i\theta} - e^{-i\theta}}{2} \tag{3} \end{aligned}
cos(θ)sin(θ)=2eiθ+e−iθ=−i⋅2eiθ−e−iθ(3)
将
s
i
n
(
θ
)
sin(\theta)
sin(θ) 、
c
o
s
(
θ
)
cos(\theta)
cos(θ) 代入傅里叶级数求得:
f
(
t
)
=
a
0
2
+
∑
n
=
1
∞
[
a
n
e
i
n
ω
t
+
e
−
i
n
ω
t
2
−
i
⋅
b
n
e
i
n
ω
t
−
e
−
i
n
ω
t
2
]
=
a
0
2
+
∑
n
=
1
∞
[
a
n
−
i
b
n
2
e
i
n
ω
t
+
a
n
+
i
b
n
2
e
−
i
n
ω
t
]
(4)
\begin{aligned} f(t)&=\frac{a_0}{2}+ \sum_{n=1}^{\infty}[a_n \frac{e^{in\omega t}+e^{-in\omega t}}{2}-i\cdot b_n \frac{e^{in\omega t}-e^{-in\omega t}}{2}]\\ &=\frac{a_0}{2}+\sum_{n=1}^{\infty}[\frac{a_n-ib_n}{2}e^{in\omega t}+\frac{a_n+ib_n}{2}e^{-in\omega t}] \tag{4} \end{aligned}
f(t)=2a0+n=1∑∞[an2einωt+e−inωt−i⋅bn2einωt−e−inωt]=2a0+n=1∑∞[2an−ibneinωt+2an+ibne−inωt](4)
将
a
0
,
a
n
,
b
n
a_0,a_n,b_n
a0,an,bn代入公式
4
{4}
4中,得到:
a
n
−
i
b
n
2
=
1
T
[
∫
t
0
t
0
+
T
f
(
t
)
c
o
s
(
n
ω
t
)
d
t
−
i
⋅
∫
t
0
t
0
+
T
f
(
t
)
s
i
n
(
n
ω
t
)
d
t
]
=
1
T
∫
t
0
t
0
+
T
f
(
t
)
[
c
o
s
(
n
ω
t
)
d
t
−
i
⋅
s
i
n
(
n
ω
t
)
]
d
t
=
1
T
∫
t
0
t
0
+
T
f
(
t
)
[
e
i
n
ω
t
+
e
−
i
n
ω
t
2
−
i
⋅
(
−
i
)
e
i
n
ω
t
−
e
−
i
n
ω
t
2
]
d
t
=
1
T
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
(5)
\begin{aligned} \frac{a_n-ib_n}{2}&=\frac{1}{T}[\int_{t_0}^{t_0+T}f(t)cos(n\omega t)dt-i\cdot\int_{t_0}^{t_0+T}f(t)sin(n\omega t)dt] \\ &=\frac{1}{T}\int_{t_0}^{t_0+T}f(t)[cos(n\omega t)dt-i\cdot sin(n\omega t)]dt \\ &=\frac{1}{T}\int_{t_0}^{t_0+T}f(t)[\frac{e^{in\omega t} + e^{-in\omega t}}{2}-i\cdot (-i)\frac{e^{in\omega t} - e^{-in\omega t}}{2}]dt \\ &=\frac{1}{T}\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt \tag{5} \end{aligned}
2an−ibn=T1[∫t0t0+Tf(t)cos(nωt)dt−i⋅∫t0t0+Tf(t)sin(nωt)dt]=T1∫t0t0+Tf(t)[cos(nωt)dt−i⋅sin(nωt)]dt=T1∫t0t0+Tf(t)[2einωt+e−inωt−i⋅(−i)2einωt−e−inωt]dt=T1∫t0t0+Tf(t)e−inωtdt(5)
同理:
a
n
+
i
b
n
2
=
1
T
∫
t
0
t
0
+
T
f
(
t
)
e
i
n
ω
t
d
t
(6)
\frac{a_n+ib_n}{2}=\frac{1}{T}\int_{t_0}^{t_0+T}f(t)e^{in\omega t} dt \tag{6}
2an+ibn=T1∫t0t0+Tf(t)einωtdt(6)
将 ( 5 ) , ( 6 ) (5),(6) (5),(6)代入 ( 4 ) (4) (4)中,得到:
f
(
t
)
=
1
T
∫
t
0
t
0
+
T
f
(
t
)
d
t
+
1
T
∑
n
=
1
∞
[
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
⋅
e
i
n
ω
t
+
∫
t
0
t
0
+
T
f
(
t
)
e
i
n
ω
t
d
t
⋅
e
−
i
n
ω
t
]
=
1
T
∫
t
0
t
0
+
T
f
(
t
)
d
t
+
1
T
∑
n
=
1
∞
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
⋅
e
i
n
ω
t
+
∑
n
=
−
∞
−
1
1
T
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
⋅
e
i
n
ω
t
=
1
T
∑
n
=
−
∞
∞
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
⋅
e
i
n
ω
t
\begin{aligned} f(t)&=\frac{1}{T}\int_{t_0}^{t_0+T}f(t)dt +\frac{1}{T}\sum_{n=1}^{\infty}[\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt\cdot e^{in\omega t}+\int_{t_0}^{t_0+T}f(t)e^{in\omega t} dt\cdot e^{-in\omega t}] \\ &=\frac{1}{T}\int_{t_0}^{t_0+T}f(t)dt +\frac{1}{T}\sum_{n=1}^{\infty}\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt\cdot e^{in\omega t}+\sum_{n=-\infty}^{-1}\frac{1}{T}\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt\cdot e^{in\omega t} \\ &=\frac{1}{T}\sum_{n=-\infty}^{\infty}\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt\cdot e^{in\omega t} \end{aligned}
f(t)=T1∫t0t0+Tf(t)dt+T1n=1∑∞[∫t0t0+Tf(t)e−inωtdt⋅einωt+∫t0t0+Tf(t)einωtdt⋅e−inωt]=T1∫t0t0+Tf(t)dt+T1n=1∑∞∫t0t0+Tf(t)e−inωtdt⋅einωt+n=−∞∑−1T1∫t0t0+Tf(t)e−inωtdt⋅einωt=T1n=−∞∑∞∫t0t0+Tf(t)e−inωtdt⋅einωt
当
n
=
0
n=0
n=0时:
1
T
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
⋅
e
i
n
ω
t
=
1
T
∫
t
0
t
0
+
T
f
(
t
)
d
t
(7)
\frac{1}{T}\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt\cdot e^{in\omega t}=\frac{1}{T}\int_{t_0}^{t_0+T}f(t)dt \tag{7}
T1∫t0t0+Tf(t)e−inωtdt⋅einωt=T1∫t0t0+Tf(t)dt(7)
接下来是与原文不同的推导
f
(
t
)
=
=
1
T
∑
n
=
−
∞
∞
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
⋅
e
i
n
ω
t
(8)
\begin{aligned} f(t)&= &=\frac{1}{T}\sum_{n=-\infty}^{\infty}\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt\cdot e^{in\omega t} \\ \tag{8} \end{aligned}
f(t)==T1n=−∞∑∞∫t0t0+Tf(t)e−inωtdt⋅einωt(8)
公式
8
8
8中,令
Ω
=
n
ω
F
(
Ω
)
=
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
Ω
t
d
t
\begin{aligned} \Omega&= n\omega\\ F(\Omega)&=\int_{t_0}^{t_0+T}f(t)e^{-i\Omega t} dt \\ \end{aligned}
ΩF(Ω)=nω=∫t0t0+Tf(t)e−iΩtdt
进一步对
F
(
Ω
)
F(\Omega)
F(Ω)进行分析:
F ( Ω ) = ∫ t 0 t 0 + T f ( t ) e − i Ω t d t = ∫ t 0 t 0 + T f ( t ) e − i ω n t d t = ∫ t 0 t 0 + T f ( t ) [ c o s ( ω n t ) − i ⋅ s i n ( ω n t ) ] d t = ∫ t 0 t 0 + T f ( t ) c o s ( ω n t ) d t − i ⋅ ∫ t 0 t 0 + T f ( t ) s i n ( ω n t ) d t = T 2 ( a n − i ⋅ b n ) = 2 π ω ⋅ 1 2 ( a n − i ⋅ b n ) \begin{aligned} F(\Omega)&=\int_{t_0}^{t_0+T}f(t)e^{-i\Omega t} dt \\ &=\int_{t_0}^{t_0+T}f(t)e^{-i\omega nt} dt \\ &=\int_{t_0}^{t_0+T}f(t)[cos(\omega nt)-i\cdot sin(\omega nt)] dt \\ &=\int_{t_0}^{t_0+T}f(t)cos(\omega nt)dt-i\cdot \int_{t_0}^{t_0+T}f(t) sin(\omega nt) dt \\ &=\frac{T}{2}( a_n-i\cdot b_n) \\ &=\frac{2\pi}{\omega}\cdot \frac{1}{2}( a_n-i\cdot b_n) \end{aligned} F(Ω)=∫t0t0+Tf(t)e−iΩtdt=∫t0t0+Tf(t)e−iωntdt=∫t0t0+Tf(t)[cos(ωnt)−i⋅sin(ωnt)]dt=∫t0t0+Tf(t)cos(ωnt)dt−i⋅∫t0t0+Tf(t)sin(ωnt)dt=2T(an−i⋅bn)=ω2π⋅21(an−i⋅bn)
由此可以看出, F ( Ω ) F(\Omega) F(Ω)的实部是余弦函数的的幅度值,虚部是正弦函数的幅度值。
公式
8
8
8的说明,因为
ω
\omega
ω很小,而,
n
n
n是从
−
∞
-\infty
−∞取到
+
∞
+\infty
+∞,因此
Ω
∈
[
−
∞
,
+
∞
]
\Omega\in[-\infty,+\infty]
Ω∈[−∞,+∞]。
则
(
8
)
(8)
(8)变为:
f
(
t
)
=
1
T
∑
n
=
−
∞
∞
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
n
ω
t
d
t
⋅
e
i
n
ω
t
=
1
T
∑
n
=
−
∞
∞
F
(
Ω
)
⋅
e
i
Ω
t
=
1
T
∫
−
∞
∞
F
(
Ω
)
⋅
e
i
Ω
t
d
n
=
1
T
ω
∫
−
∞
∞
F
(
Ω
)
⋅
e
i
Ω
t
d
ω
n
=
1
2
π
∫
−
∞
∞
F
(
Ω
)
⋅
e
i
Ω
t
d
Ω
(9)
\begin{aligned} f(t)&=\frac{1}{T}\sum_{n=-\infty}^{\infty}\int_{t_0}^{t_0+T}f(t)e^{-in\omega t} dt\cdot e^{in\omega t} \\ &=\frac{1}{T}\sum_{n=-\infty}^{\infty}F(\Omega)\cdot e^{i\Omega t} \\ &=\frac{1}{T}\int_{-\infty}^{\infty} F(\Omega)\cdot e^{i\Omega t}dn \\ &=\frac{1}{T\omega}\int_{-\infty}^{\infty} F(\Omega)\cdot e^{i\Omega t}d\omega n \\ &=\frac{1}{2\pi}\int_{-\infty}^{\infty} F(\Omega)\cdot e^{i\Omega t}d\Omega \\ \tag{9} \end{aligned}
f(t)=T1n=−∞∑∞∫t0t0+Tf(t)e−inωtdt⋅einωt=T1n=−∞∑∞F(Ω)⋅eiΩt=T1∫−∞∞F(Ω)⋅eiΩtdn=Tω1∫−∞∞F(Ω)⋅eiΩtdωn=2π1∫−∞∞F(Ω)⋅eiΩtdΩ(9)
由此,可得傅里叶变换的公式为:
f ( t ) = 1 2 π ∫ − ∞ ∞ F ( Ω ) ⋅ e i Ω t d Ω F ( Ω ) = ∫ t 0 t 0 + T f ( t ) e − i Ω t d t (10) \begin{aligned} f(t)&=\frac{1}{2\pi}\int_{-\infty}^{\infty} F(\Omega)\cdot e^{i\Omega t}d\Omega \\ F(\Omega)&=\int_{t_0}^{t_0+T}f(t)e^{-i\Omega t} dt \\ \tag{10} \end{aligned} f(t)F(Ω)=2π1∫−∞∞F(Ω)⋅eiΩtdΩ=∫t0t0+Tf(t)e−iΩtdt(10)
2、再把傅里叶的指数形式推导到离散傅里叶的形式
此部分的参考为离散傅里叶变换
在傅里叶变换的过程中,函数
f
(
t
)
f(t)
f(t)往往是连续的并且周期T是无穷大的,但是自然界中,函数
f
(
t
)
f(t)
f(t)一般是离散的,因此需要将傅里叶变换进行离散化。
f
(
t
)
f(t)
f(t)是离散的,并且个数是
N
N
N,因此假设:
f
(
1
)
=
A
1
f
(
2
)
=
A
2
f
(
3
)
=
A
3
f
(
4
)
=
A
4
f
(
5
)
=
A
5
⋮
f
(
N
)
=
A
N
\begin{aligned} f(1)&=A_ 1 \\ f(2)&=A_2 \\ f(3)&=A_3 \\ f(4)&=A_4 \\ f(5)&=A_5 \\ \vdots & \\ f(N)&=A_N \\ \end{aligned}
f(1)f(2)f(3)f(4)f(5)⋮f(N)=A1=A2=A3=A4=A5=AN
将公式
10
10
10进行离散化变形,得到
F
(
Ω
)
=
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
Ω
t
d
t
F
(
2
π
T
⋅
n
)
=
∫
t
0
t
0
+
T
f
(
t
)
e
−
i
2
π
n
t
T
d
t
G
(
n
)
=
F
(
2
π
T
⋅
n
)
=
∑
t
=
1
N
f
(
t
)
e
−
i
2
π
n
t
N
(10)
\begin{aligned} F(\Omega)&=\int_{t_0}^{t_0+T}f(t)e^{-i\Omega t} dt \\ F(\frac{2\pi}{T}\cdot n )&=\int_{t_0}^{t_0+T}f(t)e^{\frac{-i2\pi nt}{T}} dt \\ G(n)&=F(\frac{2\pi}{T}\cdot n) \\ &= \sum_{t=1}^{N}f(t)e^{\frac{-i2\pi nt}{N}} \tag{10} \end{aligned}
F(Ω)F(T2π⋅n)G(n)=∫t0t0+Tf(t)e−iΩtdt=∫t0t0+Tf(t)eT−i2πntdt=F(T2π⋅n)=t=1∑Nf(t)eN−i2πnt(10)
公式中的:
n
<
=
N
T
=
N
ω
=
2
π
T
Ω
=
2
π
T
⋅
n
\begin{aligned} n&<= N\\ T&=N \\ \omega&=\frac{2\pi}{T} \\ \Omega&=\frac{2\pi}{T} \cdot n \\ \end{aligned}
nTωΩ<=N=N=T2π=T2π⋅n
f
(
t
)
=
1
2
π
∫
−
∞
∞
F
(
Ω
)
⋅
e
i
Ω
t
d
Ω
=
1
2
π
∫
−
∞
∞
F
(
Ω
)
⋅
e
i
2
π
T
⋅
n
t
d
2
π
T
⋅
n
=
1
2
π
∫
−
∞
∞
F
(
2
π
n
T
)
⋅
e
i
2
π
N
⋅
n
t
d
2
π
N
⋅
n
=
1
2
π
⋅
2
π
N
∫
−
∞
∞
G
(
n
)
⋅
e
i
2
π
N
⋅
n
t
d
n
=
1
N
∑
n
=
1
N
G
(
n
)
⋅
e
i
2
π
N
⋅
n
t
(10)
\begin{aligned} f(t)&=\frac{1}{2\pi}\int_{-\infty}^{\infty} F(\Omega)\cdot e^{i\Omega t}d\Omega \\ &=\frac{1}{2\pi}\int_{-\infty}^{\infty} F(\Omega)\cdot e^{i \frac{2\pi}{T} \cdot n t}d\frac{2\pi}{T} \cdot n \\ &=\frac{1}{2\pi}\int_{-\infty}^{\infty} F(\frac{2\pi n}{T})\cdot e^{i \frac{2\pi}{N} \cdot n t}d\frac{2\pi}{N} \cdot n \\ &=\frac{1}{2\pi} \cdot \frac{2\pi}{N} \int_{-\infty}^{\infty} G(n)\cdot e^{i \frac{2\pi}{N} \cdot n t}d n \\ &= \frac{1}{N} \sum_{n=1}^{N} G(n)\cdot e^{i \frac{2\pi}{N} \cdot n t} \\ \tag{10} \end{aligned}
f(t)=2π1∫−∞∞F(Ω)⋅eiΩtdΩ=2π1∫−∞∞F(Ω)⋅eiT2π⋅ntdT2π⋅n=2π1∫−∞∞F(T2πn)⋅eiN2π⋅ntdN2π⋅n=2π1⋅N2π∫−∞∞G(n)⋅eiN2π⋅ntdn=N1n=1∑NG(n)⋅eiN2π⋅nt(10)
傅里叶变换后,结果是复数,其含义为:
傅氏变换后得到的复数,实部就代表该频率下的余弦信号分量,虚部就代表该频率下的正弦信号分量。
参考一维傅里叶变换后的复数怎样理解?
另一篇关于离散傅立叶变换的博客讲的很好
深入理解离散傅里叶变换(DFT)
快速傅里叶变换(FFT)算法【详解】
傅立叶变换中,为什么要引入复数
另一篇博客,从另一个角度阐述了傅立叶变换中的复数的含义
李泽光–傅立叶变换的复数理解