从傅里叶级数到傅里叶变换

定义1:傅里叶变换定义式

(1) F ( ω ) = ∫ − ∞ + ∞ f ( t ) e − j ω t d t F(\omega) = \int^{+\infty}_{-\infty} f(t) e^{-j\omega t} dt \tag{1} F(ω)=+f(t)ejωtdt(1)

定义2:傅里叶级数
对于再 [ − T / 2 , T / 2 ] [-T/2,T/2] [T/2,T/2]上,周期为 T T T的函数 f T ( t ) f_{T}(t) fT(t),有
(2) f T ( t ) ∼ a 0 2 + ∑ n = 1 + ∞ [ a n c o s ( n ω t ) + b n s i n ( n ω t ) ] f_{T}(t) \sim \frac{a_0}{2}+\sum^{+\infty}_{n=1}[a_ncos(n\omega t)+b_nsin(n\omega t)] \tag{2} fT(t)2a0+n=1+[ancos(nωt)+bnsin(nωt)](2)
其中 ω = 2 π / T \omega=2\pi/T ω=2π/T
a n = 2 T ∫ − T / 2 T / 2 f T ( t ) c o s ( n ω t ) d t a_n = \frac{2}{T}\int_{-T/2}^{T/2}f_{T}(t)cos(n\omega t)dt an=T2T/2T/2fT(t)cos(nωt)dt
b n = 2 T ∫ − T / 2 T / 2 f T ( t ) s i n ( n ω t ) d t b_n = \frac{2}{T}\int_{-T/2}^{T/2}f_{T}(t)sin(n\omega t)dt bn=T2T/2T/2fT(t)sin(nωt)dt
显然,
a 0 = 2 T ∫ − T / 2 T / 2 f T ( t ) d t a_0=\frac{2}{T}\int_{-T/2}^{T/2}f_{T}(t)dt a0=T2T/2T/2fT(t)dt

这些东西都是怎么来的呢? 下面我们分解一下

引用:https://blog.csdn.net/constantin_ouc/article/details/78881709

1.前提: 我们基于一个认识,即函数 f ( t ) f(t) f(t)可以分解为若干个余弦函数的加权和的形式。即,
(1.1) f ( t ) = A 0 c o s ( k 0 t + ψ 0 ) + A 1 c o s ( k 1 t + ψ 1 ) + ⋯ + A n c o s ( k n t + ψ n ) f(t) = A_0cos(k_0t+\psi_0)+A_1cos(k_1t+\psi_1)+\dots +A_ncos(k_nt+\psi_n) \tag{1.1} f(t)=A0cos(k0t+ψ0)+A1cos(k1t+ψ1)++Ancos(knt+ψn)(1.1)
其中, A i A_i Ai为余弦波的振幅, ψ i \psi_i ψi为对应余弦分量 A 1 c o s ( k 1 t + ψ 1 ) A_1cos(k_1t+\psi_1) A1cos(k1t+ψ1)的相位角, i = 0 , 1 , … i=0,1,\dots i=0,1,

那么,对于数轴 t t t上不同的值, t 1 , t 2 , … , t m t_1,t_2,\dots,t_m t1,t2,,tm
(1.2) { f ( t 1 ) = A 0 c o s ( k 0 t 1 + ψ 0 ) + A 1 c o s ( k 1 t 1 + ψ 1 ) + ⋯ + A n c o s ( k n t 1 + ψ n ) f ( t 2 ) = A 0 c o s ( k 0 t 2 + ψ 0 ) + A 1 c o s ( k 1 t 2 + ψ 1 ) + ⋯ + A n c o s ( k n t 2 + ψ n ) … f ( t m ) = A 0 c o s ( k 0 t m + ψ 0 ) + A 1 c o s ( k 1 t m + ψ 1 ) + ⋯ + A n c o s ( k n t m + ψ n ) \begin{cases} f(t_1) = A_0cos(k_0t_1+\psi_0)+A_1cos(k_1t_1+\psi_1)+\dots +A_ncos(k_nt_1+\psi_n) \\ f(t_2) = A_0cos(k_0t_2+\psi_0)+A_1cos(k_1t_2+\psi_1)+\dots +A_ncos(k_nt_2 +\psi_n) \\ \dots \\ f(t_m) = A_0cos(k_0t_m+\psi_0)+A_1cos(k_1t_m+\psi_1)+\dots +A_ncos(k_nt_m +\psi_n) \tag{1.2} \end{cases} f(t1)=A0cos(k0t1+ψ0)+A1cos(k1t1+ψ1)++Ancos(knt1+ψn)f(t2)=A0cos(k0t2+ψ0)+A1cos(k1t2+ψ1)++Ancos(knt2+ψn)f(tm)=A0cos(k0tm+ψ0)+A1cos(k1tm+ψ1)++Ancos(kntm+ψn)(1.2)

A = [ c o s ( k 0 t 1 + ψ 0 ) c o s ( k 1 t 1 + ψ 1 ) … c o s ( k n t 1 + ψ n ) c o s ( k 0 t 2 + ψ 0 ) c o s ( k 1 t 2 + ψ 1 ) … c o s ( k n t 1 + ψ n ) … … … … c o s ( k 0 t m + ψ 0 ) c o s ( k 1 t m + ψ 1 ) … c o s ( k n t m + ψ n ) ] A = \begin{bmatrix} cos(k_0t_1+\psi_0) && cos(k_1t_1+\psi_1)&&\dots &&cos(k_nt_1+\psi_n) \\ cos(k_0t_2+\psi_0) && cos(k_1t_2+\psi_1)&&\dots &&cos(k_nt_1+\psi_n) \\ \dots && \dots && \dots&& \dots\\ cos(k_0t_m+\psi_0) && cos(k_1t_m+\psi_1)&&\dots &&cos(k_nt_m+\psi_n) \end{bmatrix} A=cos(k0t1+ψ0)cos(k0t2+ψ0)cos(k0tm+ψ0)cos(k1t1+ψ1)cos(k1t2+ψ1)cos(k1tm+ψ1)cos(knt1+ψn)cos(knt1+ψn)cos(kntm+ψn),
X = [ a 0 a 1 … a n ] T X = \begin{bmatrix} a_0 && a_1 &&\dots && a_n \end{bmatrix}^{\rm{T}} X=[a0a1an]T,
B = [ f ( t 1 ) f ( t 2 ) … f ( t m ) ] T B = \begin{bmatrix} f(t_1) && f(t_2) &&\dots && f(t_m) \end{bmatrix}^{\rm{T}} B=[f(t1)f(t2)f(tm)]T
得到 B = A X B=AX B=AX
X = A − 1 B X=A^{-1}B X=A1B可以得到系数向量 X X X

在(1.1)中, f ( t ) = A 0 c o s ( k 0 t + ψ 0 ) + A 1 c o s ( k 1 t + ψ 1 ) + ⋯ + A n c o s ( k n t + ψ n ) f(t) = A_0cos(k_0t+\psi_0)+A_1cos(k_1t+\psi_1)+\dots +A_ncos(k_nt+\psi_n) f(t)=A0cos(k0t+ψ0)+A1cos(k1t+ψ1)++Ancos(knt+ψn)的角频率 { k i } \{k_i\} {ki}为随意选取的值。显然,不同的角频率 { k i } \{k_i\} {ki}可以得到不同的余弦分量系数 { a i } \{a_i\} {ai}

为了便于处理,我们规定角频率 { k i } \{k_i\} {ki}的选取。令 k i = i , ( i = 0 , 1 , …   ) k_i=i,(i=0,1,\dots ) ki=i,(i=0,1,)

我们看到 f ( t ) = ∑ i = 0 n A i c o s ( k i t + ψ i ) f(t)=\sum_{i=0}^{n} A_icos(k_it+\psi_i) f(t)=i=0nAicos(kit+ψi)中,相位角 ψ i \psi_i ψi 太讨厌了! 不利于分析!我们把相位角处理掉
f ( t ) = ∑ i = 0 n A i c o s ( k i t + ψ i ) = ∑ i = 0 n A i ( c o s ( k i t ) c o s ( ψ i ) − s i n ( k i t ) s i n ( ψ i ) ) = ∑ i = 0 n A i c o s ( ψ i ) c o s ( k i t ) − A i s i n ( ψ i ) s i n ( k i t ) f(t)=\sum_{i=0}^{n} A_icos(k_it+\psi_i)\\ =\sum_{i=0}^{n}A_i(cos(k_it)cos(\psi_i)-sin(k_it)sin(\psi_i))\\ =\sum_{i=0}^{n}A_icos(\psi_i)cos(k_it)-A_isin(\psi_i)sin(k_it) f(t)=i=0nAicos(kit+ψi)=i=0nAi(cos(kit)cos(ψi)sin(kit)sin(ψi))=i=0nAicos(ψi)cos(kit)Aisin(ψi)sin(kit)
a i = − A i s i n ( ψ i ) , b i = A i c o s ( ψ i ) a_i =- A_isin(\psi_i),b_i = A_icos(\psi_i) ai=Aisin(ψi),bi=Aicos(ψi),则上式有

(1.3) f ( t ) = ∑ i = 0 n [ a i s i n ( k i t ) + b i c o s ( k i t ) ] f(t)=\sum_{i=0}^{n}[a_isin(k_it)+b_icos(k_it)] \tag{1.3} f(t)=i=0n[aisin(kit)+bicos(kit)](1.3)
以上为傅里叶级数的展开形式。
\\ \\
这样做有什么好处呢?
答:
好处1(周期性):对于(1.3)中 f ( t ) = ∑ i = 0 n [ a i s i n ( k i t ) + b i c o s ( k i t ) ] f(t)=\sum_{i=0}^{n}[a_isin(k_it)+b_icos(k_it)] f(t)=i=0n[aisin(kit)+bicos(kit)]而言,记 T i T_i Ti为期周期,则
T i = 2 π k i T_i = \frac{2\pi}{k_i} Ti=ki2π
由于 k i = i , ( i = 0 , 1 , …   ) k_i=i,(i=0,1,\dots ) ki=i,(i=0,1,),使得 { T i } \{T_i\} {Ti}有公共的周期 T = 2 π T= 2\pi T=2π,这使得 f ( t ) f(t) f(t)成为周期为 2 π 2\pi 2π的函数。

好处2:正弦函数、余弦函数在一个周期内的积分等于0
(1.4) ∫ − T 2 T 2 a i c o s ( k i t ) d t = ∫ 0 T a i c o s ( k i t ) d t = 0 \int_{-\frac{T}{2}}^{\frac{T}{2}} a_icos(k_it) dt =\int_{0}^{T} a_icos(k_it) dt= 0 \tag{1.4} 2T2Taicos(kit)dt=0Taicos(kit)dt=0(1.4)

\\
好处说完了,但是(1.3)中 f ( t ) = ∑ i = 0 n [ a i s i n ( k i t ) + b i c o s ( k i t ) ] f(t)=\sum_{i=0}^{n}[a_isin(k_it)+b_icos(k_it)] f(t)=i=0n[aisin(kit)+bicos(kit)]中的系数 a i , b i a_i,b_i ai,bi怎么求呢?
鉴于 f ( t ) f(t) f(t)有周期 T = 2 π T=2\pi T=2π,我们设计一个积分形式

(1.5) ∫ − π π f ( t ) c o s ( k i t ) d t = ∫ − π π ∑ j = 0 n [ a j s i n ( k j t ) + b j c o s ( k j t ) ] c o s ( k i t ) d t = ∫ − π π ∑ j = 0 n [ a j s i n ( k j t ) c o s ( k i t ) + b j c o s ( k j t ) c o s ( k i t ) ] d t = ∫ − π π { b 0 c o s ( k i t ) + ∑ j = 1 n [ a j s i n ( k j t ) c o s ( k i t ) + b j c o s ( k j t ) c o s ( k i t ) ] } d t = 2 π b 0 c o s ( k i t ) + ∫ − π π ∑ j = 1 n [ a j s i n ( k j t ) c o s ( k i t ) ] d t + ∫ − π π ∑ j = 1 n [ b j c o s ( k j t ) c o s ( k i t ) ] d t \int_{-\pi}^{\pi}f(t)cos(k_it)dt=\int_{-\pi}^{\pi}\sum_{j=0}^{n}[a_jsin(k_jt)+b_jcos(k_jt)]cos(k_it)dt\\ =\int_{-\pi}^{\pi}\sum_{j=0}^{n}[a_jsin(k_jt)cos(k_it)+b_jcos(k_jt)cos(k_it)]dt\\ =\int_{-\pi}^{\pi} \{b_0cos(k_it)+\sum_{j=1}^{n}[a_jsin(k_jt)cos(k_it)+b_jcos(k_jt)cos(k_it)]\}dt\\ =2\pi b_0cos(k_it)+\int_{-\pi}^{\pi}\sum_{j=1}^{n}[a_jsin(k_jt)cos(k_it)]dt+\int_{-\pi}^{\pi}\sum_{j=1}^{n}[b_jcos(k_jt)cos(k_it)]dt \tag{1.5}\\ ππf(t)cos(kit)dt=ππj=0n[ajsin(kjt)+bjcos(kjt)]cos(kit)dt=ππj=0n[ajsin(kjt)cos(kit)+bjcos(kjt)cos(kit)]dt=ππ{b0cos(kit)+j=1n[ajsin(kjt)cos(kit)+bjcos(kjt)cos(kit)]}dt=2πb0cos(kit)+ππj=1n[ajsin(kjt)cos(kit)]dt+ππj=1n[bjcos(kjt)cos(kit)]dt(1.5)
上式中第一项,正、余弦函数在一个周期内的积分为0,故第一项为0;第二项为奇函数积分,为0;第三项,根据三角函数的正交性,当 k j ≠ k i k_j\neq k_i kj̸=ki时,为0;仅当 k j = k i k_j=k_i kj=ki时不为0。

故,第三项为
(1.6) ∫ − π π ∑ j = 1 n [ b j c o s ( k j t ) c o s ( k i t ) ] d t = ∫ − π π [ b i c o s 2 ( k i t ) ] d t = b i 2 ∫ − π π ( 1 − c o s ( 2 k i t ) ) d t ( 半 角 公 式 ) = π b i − 0 = π b i \int_{-\pi}^{\pi}\sum_{j=1}^{n}[b_jcos(k_jt)cos(k_it)]dt\\ =\int_{-\pi}^{\pi}[b_icos^2(k_it)]dt\\ =\frac{b_i}{2}\int_{-\pi}^{\pi}(1-cos(2k_it))dt (半角公式)\\ =\pi b_i-0 = \pi b_i \tag{1.6} ππj=1n[bjcos(kjt)cos(kit)]dt=ππ[bicos2(kit)]dt=2biππ(1cos(2kit))dt=πbi0=πbi(1.6)

结合(1.5)、(1.6), ∫ − π π f ( t ) c o s ( k i t ) d t = π b i \int_{-\pi}^{\pi}f(t)cos(k_it)dt=\pi b_i ππf(t)cos(kit)dt=πbi,可得
(1.7) b i = 1 π ∫ − π π f ( t ) c o s ( k i t ) d t b_i = \frac{1}{\pi}\int_{-\pi}^{\pi}f(t)cos(k_it)dt \tag{1.7} bi=π1ππf(t)cos(kit)dt(1.7)

同理,与(1.5)一致,取 ∫ − π π f ( t ) s i n ( k i t ) d t \int_{-\pi}^{\pi}f(t)sin(k_it)dt ππf(t)sin(kit)dt,可得
(1.8) a i = 1 π ∫ − π π f ( t ) s i n ( k i t ) d t a_i = \frac{1}{\pi}\int_{-\pi}^{\pi}f(t)sin(k_it)dt \tag{1.8} ai=π1ππf(t)sin(kit)dt(1.8)

以上就是傅里叶级数的由来

下面我们要推出傅里叶变换

傅 里 叶 级 数 ⟶ e i x = c o s x + i s i n x 傅 里 叶 变 换 傅里叶级数 \stackrel{e^{ix}=cosx + isinx}{\longrightarrow} 傅里叶变换 eix=cosx+isinx

j 2 = − 1 j^2=-1 j2=1

(1.9) c o s θ = e j θ + e − j θ 2 s i n θ = e j θ − e − j θ 2 = − j e j θ − e − j θ 2 cos\theta = \frac{e^{j\theta}+e^{-j\theta}}{2} \\ sin\theta = \frac{e^{j\theta}-e^{-j\theta}}{2}=-j\frac{e^{j\theta}-e^{-j\theta}}{2} \tag{1.9} cosθ=2ejθ+ejθsinθ=2ejθejθ=j2ejθejθ(1.9)
(1.3)写成
f T ( x ) = a 0 2 + ∑ n = 1 ∞ ( a n s i n n ω x + b n c o s n ω x ) = a 0 2 + ∑ n = 1 ∞ ( − j a n e j n ω x − e − j n ω x 2 + b n e j n ω x + e − j n ω x 2 ) = a 0 2 + ∑ n = 1 ∞ ( b n − j a n 2 e j n ω x + b n − j a n 2 e − j n ω x ) f_T(x) = \frac{a_0}{2}+\sum_{n=1}^{\infty}(a_nsinn\omega x+b_ncosn\omega x) \\ = \frac{a_0}{2}+\sum_{n=1}^{\infty}( -ja_n\frac{e^{jn\omega x}-e^{-jn\omega x}}{2} +b_n \frac{e^{jn\omega x}+e^{-jn\omega x}}{2} )\\ = \frac{a_0}{2}+\sum_{n=1}^{\infty}( \frac{b_n-ja_n}{2} {e^{jn\omega x}}+ \frac{b_n-ja_n}{2} {e^{-jn\omega x}} )\\ fT(x)=2a0+n=1(ansinnωx+bncosnωx)=2a0+n=1(jan2ejnωxejnωx+bn2ejnωx+ejnωx)=2a0+n=1(2bnjanejnωx+2bnjanejnωx)


c 0 = a 0 2 = 1 T ∫ − T 2 T 2 f T ( x ) d x c_0 = \frac{a_0}{2} = \frac{1}{T}\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x)dx c0=2a0=T12T2TfT(x)dx

c n = a n − j b n 2 = 1 T [ ∫ − T 2 T 2 f T ( x ) c o s n ω x d x − j ∫ − T 2 T 2 f T ( x ) s i n n ω x d x ] = 1 T ∫ − T 2 T 2 f T ( x ) ( c o s n ω x − j s i n n ω x ) d x = 1 T ∫ − T 2 T 2 f T ( x ) e − j n ω x d x c_n = \frac{a_n -jb_n}{2} = \frac{1}{T}[\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x)cosn\omega xdx- j\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x)sinn\omega xdx] \\ =\frac{1}{T}\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x)(cosn\omega x-jsinn\omega x)dx \\ =\frac{1}{T}\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x) e^{-jn\omega x}dx cn=2anjbn=T1[2T2TfT(x)cosnωxdxj2T2TfT(x)sinnωxdx]=T12T2TfT(x)(cosnωxjsinnωx)dx=T12T2TfT(x)ejnωxdx
其中 n = ± 1 , ± 2 , … , n=\pm1,\pm2,\dots, n=±1,±2,,
若令 w n = n ω w_n=n\omega wn=nω,则
(1.10) f T ( x ) = c 0 + ∑ n = 1 ∞ ( c n e j w n x + c − n e − j w n x ) = ∑ n = − ∞ ∞ c n e j w n x = ∑ n = − ∞ ∞ [ 1 T ∫ − T 2 T 2 f T ( x ) e − j w n x d x ] e j w n x = 1 T ∑ n = − ∞ ∞ [ ∫ − T 2 T 2 f T ( x ) e − j w n x d x ] e j w n x f_T(x) =c_0 + \sum_{n=1}^{\infty}(c_ne^{jw_nx}+c_{-n}e^{-jw_nx})=\sum_{n=-\infty}^{\infty}c_ne^{jw_nx} =\sum_{n=-\infty}^{\infty} [\frac{1}{T}\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x) e^{-j w_n x}dx]e^{jw_nx}\\ = \frac{1}{T}\sum_{n=-\infty}^{\infty} [\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x) e^{-j w_n x}dx]e^{jw_nx} \tag{1.10} fT(x)=c0+n=1(cnejwnx+cnejwnx)=n=cnejwnx=n=[T12T2TfT(x)ejwnxdx]ejwnx=T1n=[2T2TfT(x)ejwnxdx]ejwnx(1.10)
我们来解读下(1.10)

f T ( x ) = 1 T ∑ n = − ∞ ∞ [ ∫ − T 2 T 2 f T ( x ) e − j w n x d x ] e j w n x = 1 T ∑ n = − ∞ ∞ [ F ( w n ) ] e j w n x f_T(x)= \frac{1}{T}\sum_{n=-\infty}^{\infty} [\int_{-\frac{T}{2}}^{\frac{T}{2}}f_T(x) e^{-j w_n x}dx]e^{jw_nx} \\ =\frac{1}{T}\sum_{n=-\infty}^{\infty} [F(w_n)]e^{jw_nx} fT(x)=T1n=[2T2TfT(x)ejwnxdx]ejwnx=T1n=[F(wn)]ejwnx
上式表明,周期函数 f T ( x ) f_T(x) fT(x)由幅值 F ( w n ) F(w_n) F(wn)和频率 e j w n x e^{jw_nx} ejwnx的累加和组成。

评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
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、付费专栏及课程。

余额充值