在第9部分学习笔记中,我们导出了分布傅里叶变换,用以解决普通傅里叶变换难以解决的一些问题。本文的内容就是在此基础上进一步进行讨论。
一、分布傅里叶变换的导数相关性质
< T ′ , φ > = ∫ − ∞ ∞ T ′ ( x ) φ ( x ) d x = T ( x ) φ ( x ) ∣ − ∞ ∞ − ∫ − ∞ ∞ T ( x ) d φ ( x ) \begin{aligned} \left< T^{'}, \varphi \right> &=\int_{-\infty}^{\infty}T^{'}(x)\varphi(x)dx\\ &=\left. T(x)\varphi(x) \right|_{-\infty}^{\infty} - \int_{-\infty}^{\infty}T(x) d\varphi(x)\\ \end{aligned} ⟨T′,φ⟩=∫−∞∞T′(x)φ(x)dx=T(x)φ(x)∣−∞∞−∫−∞∞T(x)dφ(x)
由于是缓增分布,是速降函数,因此当 x → ± ∞ x \to \pm \infty x→±∞时, φ ( x ) → 0 \varphi(x) \to 0 φ(x)→0, T ( x ) φ ( x ) → 0 T(x)\varphi(x) \to 0 T(x)φ(x)→0,于是就有:
< T ′ , φ > = 0 − ∫ − ∞ ∞ T ( x ) φ ′ ( x ) d x = − < T , φ ′ > \begin{aligned} \left< T^{'}, \varphi \right> &=0-\int_{-\infty}^{\infty}T(x)\varphi^{'}(x)dx\\ &= -\left< T, \varphi^{'} \right> \end{aligned} ⟨T′,φ⟩=0−∫−∞∞T(x)φ′(x)dx=−⟨T,φ′⟩
这就是我们推导的第一个非常重要的公式:
< T ′ , φ > = − < T , φ ′ > (1) \huge\left< T^{'}, \varphi \right> = -\left< T, \varphi^{'} \right>\tag 1 ⟨T′,φ⟩=−⟨T,φ′⟩(1)
接下来我们举几个例子来应用上述公式,
-
单位阶跃函数 u ( x ) u(x) u(x)
u ( x ) = { 1 x > 0 1 2 x = 0 0 x < 0 u(x)= \begin{cases} 1 & \quad x>0\\ \frac{1}{2} & \quad x=0\\ 0 & \quad x<0 \end{cases} u(x)=⎩⎪⎨⎪⎧1210x>0x=0x<0
根据 ( 1 ) (1) (1) 式,我们来求它的导数
u ′ = < u ′ , φ > = − < u , φ ′ > = − ∫ − ∞ ∞ u ( x ) φ ′ ( x ) d x = − ∫ 0 ∞ φ ′ ( x ) d x = − ( φ ( ∞ ) − φ ( 0 ) ) = φ ( 0 ) = < δ , φ > = δ \begin{aligned} u^{'}=\left< u^{'}, \varphi \right>= -\left< u, \varphi^{'} \right> &= -\int_{-\infty}^{\infty}u(x)\varphi^{'}(x)dx\\ &=-\int_{0}^{\infty} \varphi^{'}(x)dx\\ &=-\left( \varphi(\infty)-\varphi(0)\right)\\ &=\varphi(0)\\ &=\left< \delta, \varphi \right>\\ &=\delta \end{aligned} u′=⟨u′,φ⟩=−⟨u,φ′⟩=−∫−∞∞u(x)φ′(x)dx=−∫0∞φ′(x)dx=−(φ(∞)−φ(0))=φ(0)=⟨δ,φ⟩=δ
于是我们就得到了 u ′ = δ u^{'} = \delta u′=δ -
信号函数 s g n ( x ) sgn(x) sgn(x)
s g n ( x ) = { 1 x > 0 0 x = 0 − 1 x < 0 sgn(x)= \begin{cases} 1 & \quad x>0\\ 0 & \quad x=0\\ -1 & \quad x<0 \end{cases} sgn(x)=⎩⎪⎨⎪⎧10−1x>0x=0x<0
类似地,我们有
s g n ′ = < s g n ′ , φ > = − < s g n , φ ′ > = − ∫ − ∞ ∞ s g n ( x ) φ ′ ( x ) d x = − ( ∫ 0 ∞ φ ′ ( x ) d x + ∫ − ∞ 0 − φ ′ ( x ) d x ) = − [ ( φ ( ∞ ) − φ ( 0 ) ) + ( − φ ( 0 ) + φ ( − ∞ ) ) ] = 2 φ ( 0 ) = 2 < δ , φ > = 2 δ \begin{aligned} sgn^{'}&=\left< sgn^{'}, \varphi \right>\\ &= -\left< sgn, \varphi^{'} \right> \\ &= -\int_{-\infty}^{\infty}sgn(x)\varphi^{'}(x)dx\\ &=-\left( \int_{0}^{\infty} \varphi^{'}(x)dx + \int_{-\infty}^{0} -\varphi^{'}(x)dx \right)\\ &=-\left[ \left( \varphi(\infty)-\varphi(0)\right) + \left( -\varphi(0) + \varphi(-\infty)\right) \right]\\ &=2\varphi(0)\\ &=2\left< \delta, \varphi \right>\\ &=2\delta \end{aligned} sgn′=⟨sgn′,φ⟩=−⟨sgn,φ′⟩=−∫−∞∞sgn(x)φ′(x)dx=−(∫0∞φ′(x)dx+∫−∞0−φ′(x)dx)=−[(φ(∞)−φ(0))+(−φ(0)+φ(−∞))]=2φ(0)=2⟨δ,φ⟩=2δ
即 s g n ′ = 2 δ sgn^{'} = 2\delta sgn′=2δ
二、分布傅里叶导数定理
类似普通傅里叶变换中讨论的,我们有:
F ( T ( n ) ) = ( 2 π i s ) n F T \mathscr{F}\left(T^{(n)}\right) = \left( 2\pi is\right)^n \mathscr{F}T F(T(n))=(2πis)nFT
( F T ) ( n ) = F ( ( − 2 π i t ) n T ) \left( \mathscr{F} T \right) ^{(n)} = \mathscr{F} \left( \left( -2\pi it \right)^n T \right) (FT)(n)=F((−2πit)nT)
例
-
信号函数 s g n ( x ) sgn(x) sgn(x)
我们要求其傅里叶变换,首先,我们有:
F ( s g n ′ ) = F ( 2 δ ) = 2 \mathscr{F}\left(sgn^{'} \right) = \mathscr{F} (2\delta) = 2 F(sgn′)=F(2δ)=2
然后根据导数定理,有
F ( s g n ′ ) = 2 π i s F ( s g n ) \mathscr{F}\left(sgn^{'} \right) = 2\pi is \mathscr{F} \left( sgn\right) F(sgn′)=2πisF(sgn)
联立二式可以得到
F ( s g n ) = 1 π i s \mathscr{F} \left( sgn\right) = \frac{1}{\pi is} F(sgn)=πis1 -
单位阶跃函数 u ( x ) u(x) u(x)
简单来说,我们可以将 u ( x ) u(x) u(x)写成 1 2 ( 1 + s n g ( x ) ) \frac{1}{2}\left( 1+ sng(x)\right) 21(1+sng(x)),于是就有
F u = 1 2 F ( 1 + s n g ) = 1 2 ( δ + 1 π i s ) \begin{aligned} \mathscr{F}u &= \frac{1}{2} \mathscr{F} \left( 1+ sng\right)\\ &=\frac{1}{2} \left( \delta + \frac{1}{\pi is} \right) \end{aligned} Fu=21F(1+sng)=21(δ+πis1)
三、函数(分布)与分布的乘积与卷积
注意分布与分布的乘积、卷积很多情况下是没有意义的,但是函数与分布的乘积与卷积大部分情况下是有意义的
-
函数与分布的乘积公式
< f T , φ > = < T , f φ > \huge\left< fT, \varphi \right> = \left< T, f\varphi \right> ⟨fT,φ⟩=⟨T,fφ⟩
证明如下
< f T , φ > = ∫ − ∞ ∞ f ( x ) T ( x ) φ ( x ) d x = ∫ − ∞ ∞ T ( x ) ( f ( x ) φ ( x ) ) d x = < T , f φ > \begin{aligned} \left< fT, \varphi \right> &= \int_{-\infty}^{\infty}f(x)T(x)\varphi (x)dx\\ &= \int_{-\infty}^{\infty}T(x) \left(f(x)\varphi (x) \right)dx\\ &=\left< T, f\varphi \right> \end{aligned} ⟨fT,φ⟩=∫−∞∞f(x)T(x)φ(x)dx=∫−∞∞T(x)(f(x)φ(x))dx=⟨T,fφ⟩
两个常见例子:
a) f δ = f ( 0 ) δ f\delta = f(0)\delta fδ=f(0)δ
f δ = < f δ , φ > = < δ , f φ > = f ( 0 ) φ ( 0 ) = f ( 0 ) < δ , φ > = < f ( 0 ) δ , φ > = f ( 0 ) δ \begin{aligned} f\delta = \left< f\delta, \varphi \right> &= \left< \delta, f\varphi \right>\\ &= f(0) \varphi (0)\\ &=f(0)\left< \delta, \varphi \right>\\ &=\left< f(0)\delta, \varphi \right>\\ &=f(0)\delta \end{aligned} fδ=⟨fδ,φ⟩=⟨δ,fφ⟩=f(0)φ(0)=f(0)⟨δ,φ⟩=⟨f(0)δ,φ⟩=f(0)δ
b) f δ a = f ( a ) δ a \huge f\delta_a = f(a)\delta_a fδa=f(a)δa
上式被称为 δ a \delta_a δa的抽样特性,是实际应用中一个非常常用的公式 -
函数与分布的卷积公式
与普通傅里叶变换类似地,我们有:
F ( f ∗ T ) = ( F f ) ( F T ) \mathscr{F}( f*T ) = \left(\mathscr{F}f \right)\left(\mathscr{F}T \right) F(f∗T)=(Ff)(FT)
同样,我们举出一个非常常用的实例
a) F ( f ∗ δ ) = ( F f ) ( F δ ) = F f \mathscr{F}( f*\delta ) = \left(\mathscr{F}f \right)\left(\mathscr{F}\delta \right) =\mathscr{F}f F(f∗δ)=(Ff)(Fδ)=Ff
对等式两边再进行傅里叶逆变换,就有 f ∗ δ = f f*\delta =f f∗δ=f,更一般地,我们有:
( f ∗ δ a ) ( x ) = f ( x − a ) (f*\delta_a) (x) = f(x-a) (f∗δa)(x)=f(x−a)
- 分布与分布的卷积公式特例
一般来说分布与分布的卷积是没有意义的,但是也有例外,比如有一个非常重要的例子:
δ a ∗ δ b = δ a + b \delta_a * \delta_b = \delta_{a+b} δa∗δb=δa+b
四、 δ ( a x ) \delta(ax) δ(ax)
从便于计算的角度出发推导,不必刻意寻求物理意义,我们可以得到
δ ( k x ) = 1 ∣ k ∣ δ ( x ) \huge\delta(kx) = \frac{1}{|k|} \delta(x) δ(kx)=∣k∣1δ(x)
证明如下
δ ( k x ) = < δ ( k x ) , φ > = ∫ − ∞ ∞ δ ( k x ) φ ( x ) d x \delta(kx) = \left< \delta(kx), \varphi \right>=\int_{-\infty}^{\infty} \delta(kx) \varphi(x)dx δ(kx)=⟨δ(kx),φ⟩=∫−∞∞δ(kx)φ(x)dx
令 u = k x u=kx u=kx,讨论 k ≥ 0 k \geq 0 k≥0 的情况,有
δ ( k x ) = ∫ − ∞ ∞ δ ( u ) φ ( u k ) d ( u k ) = 1 k < δ ( u ) , φ ( u k ) > = 1 k φ ( 0 ) = 1 k < δ , φ > = 1 k δ ( k x ) \begin{aligned} \delta(kx)&=\int_{-\infty}^{\infty} \delta(u) \varphi(\frac{u}{k})d\left( \frac{u}{k}\right)\\ &=\frac{1}{k} \left< \delta(u), \varphi(\frac{u}{k}) \right>\\ &=\frac{1}{k} \varphi(0)\\ &=\frac{1}{k} \left< \delta, \varphi \right>\\ &=\frac{1}{k} \delta(kx) \end{aligned} δ(kx)=∫−∞∞δ(u)φ(ku)d(ku)=k1⟨δ(u),φ(ku)⟩=k1φ(0)=k1⟨δ,φ⟩=k1δ(kx)
同理可得 k < 0 k < 0 k<0 的情况,得证。