Catalogue
Basic knowledge
1.1 How to describe a wave
- [picture from Baidu]
- [formula]
x
=
A
s
i
n
(
ω
t
+
φ
)
x = Asin( ωt + φ )
x=Asin(ωt+φ)
- A means amplitude.
- ω represents angular frequency, which means phase angle radian (相角弧度值) changing per unit time. ω = 2 π f = 2 π T \omega = 2 \pi f = {2\pi \above{1pt} T} ω=2πf=T2π
- φ means primary phase, while ω t + φ ωt + φ ωt+φ means phase.
1.2 Standard orthonormal basis
- The standard orthogonal basis can represent any vector in the vector space.
1.3 Hilbert space
- Hilbert space extends Euclid space, and the base (基底) of Hilbert space is usually function. For example, arbitrary function in Hilbert space can decomposed into sine and cosine function thought Fourier Series, where sine and cosine are orthonormal basis.
1.4 Dirichlet’s convergence theorem
Let f ( x ) f(x) f(x) takes 2 l 2l 2l as a period, if f ( x ) f(x) f(x) on [ − l , l ] [-l, l] [−l,l] satisfies that:
- ①. f ( x ) f(x) f(x) continuous or only finite first class discontinuities (第一类间断点:可去、跳跃)
- ②. only finite extreme point (极值点)
we can say that the Fourier Series of f ( x ) f(x) f(x) converge in [ − l , l ] [-l, l] [−l,l].
1.5 Euler’s formula
e i x = c o s x + i ∗ s i n x e^{ix} = cosx + i*sinx eix=cosx+i∗sinx
Fourier Series
1.1 Formula
f
(
x
)
=
a
0
2
+
∑
n
=
1
∞
(
a
n
∗
c
o
s
(
n
π
x
l
)
+
b
n
∗
s
i
n
(
n
π
x
l
)
)
f(x) = {a_0 \above{1pt} 2} + \displaystyle\sum_{n=1}^\infin (a_n*cos({n\pi x \above{1pt} l}) + b_n * sin({n\pi x \above{1pt} l}))
f(x)=2a0+n=1∑∞(an∗cos(lnπx)+bn∗sin(lnπx))
where
a
0
=
1
l
∫
−
l
l
f
(
x
)
d
x
a_0 = {1 \above{1pt} l} \int_{-l}^l f(x)dx
a0=l1∫−llf(x)dx
a
n
=
1
l
∫
−
l
l
f
(
x
)
∗
c
o
s
(
n
π
x
l
)
d
x
a_n = {1 \above{1pt} l} \int_{-l}^l f(x)*cos({n\pi x \above{1pt} l})dx
an=l1∫−llf(x)∗cos(lnπx)dx
b
n
=
1
l
∫
−
l
l
f
(
x
)
∗
s
i
n
(
n
π
x
l
)
d
x
b_n = {1 \above{1pt} l} \int_{-l}^l f(x)*sin({n\pi x \above{1pt} l})dx
bn=l1∫−llf(x)∗sin(lnπx)dx
1.2 Compact formula
f
(
x
)
=
∑
k
=
−
∞
∞
c
k
∗
e
i
k
x
f(x) = \displaystyle\sum_{k=-\infin}^\infin c_k * e^{ikx}
f(x)=k=−∞∑∞ck∗eikx
where
c
k
c_k
ck represents the projection of
f
(
x
)
f(x)
f(x) in
e
i
k
x
e^{ikx}
eikx direction can be written as
<
f
(
x
)
,
e
i
k
x
>
<f(x), e^{ikx}>
<f(x),eikx>. Actually,
c
k
c_k
ck can be considered as coefficient.
Fourier Transform
1.1 Formula
- forward (time domain to frequency domain)
F ( ξ ) = ∫ − ∞ ∞ f ( t ) ∗ e − i ω t d t F(\xi) = \int_{-\infin}^\infin f(t)*e^{-i\omega t} dt F(ξ)=∫−∞∞f(t)∗e−iωtdt
where ξ \xi ξ means frequency. - inverse (frequency domain to time domain)
f ( t ) = ∫ − ∞ ∞ F ( ξ ) ∗ e i ω t d ξ f(t) = \int_{-\infin}^\infin F(\xi)*e^{i\omega t} d\xi f(t)=∫−∞∞F(ξ)∗eiωtdξ
Diff between Fourier Series & Fourier Transform
- Fourier Series is about period function, while Fourier Transform extend l l l to infinity.
Discrete Fourier Transform (DFT)
1.1 Tips
- DFT approximating that Fourier series approximation on a finite interval where your function is periodic.
1.2 Transform Formula
f
k
^
=
∑
j
=
0
n
−
1
f
j
∗
e
−
i
2
π
j
k
n
\hat{f_k} = \sum_{j=0}^{n-1} f_j * e^{-i2\pi jk \above{1pt} n}
fk^=j=0∑n−1fj∗en−i2πjk
f
k
=
1
n
∑
j
=
0
n
−
1
f
j
^
∗
e
i
2
π
j
k
n
{f_k} = {1 \above{1pt} n} \sum_{j=0}^{n-1} \hat{f_j} * e^{i2\pi jk \above{1pt} n}
fk=n1j=0∑n−1fj^∗eni2πjk
let ω = e − i 2 π n \omega = e^{-i2\pi \above{1pt} n} ω=en−i2π
[ f 0 ^ f 1 ^ f 2 ^ . . . f n ^ ] = [ 1 1 1 . . . 1 1 ω ω 2 . . . ω n − 1 1 ω 2 ω 4 . . . ω 2 ( n − 1 ) 1 . . . . . . . . . . . . 1 ω n − 1 ω 2 ( n − 1 ) . . . ω ( n − 1 ) ( n − 1 ) ] [ f 0 f 1 f 2 . . . f n ] \begin{bmatrix} \hat{f_0} \\ \hat{f_1} \\ \hat{f_2} \\ ...\\ \hat{f_n} \\ \end{bmatrix} = \begin{bmatrix} 1 & 1 & 1 & ... & 1\\ 1 & \omega & \omega^2 & ... & \omega^{n-1} \\ 1 & \omega^2 & \omega^4 & ... & \omega^{2(n-1)} \\ 1 & ... & ... & ... & ...\\ 1 & \omega^{n-1} & \omega^{2(n-1)} & ... & \omega^{(n-1)(n-1)}\\ \end{bmatrix} \begin{bmatrix} f_0 \\ f_1 \\ f_2 \\ ...\\ f_n \\ \end{bmatrix} ⎣⎢⎢⎢⎢⎡f0^f1^f2^...fn^⎦⎥⎥⎥⎥⎤=⎣⎢⎢⎢⎢⎡111111ωω2...ωn−11ω2ω4...ω2(n−1)...............1ωn−1ω2(n−1)...ω(n−1)(n−1)⎦⎥⎥⎥⎥⎤⎣⎢⎢⎢⎢⎡f0f1f2...fn⎦⎥⎥⎥⎥⎤
Fast Fourier Transform (FFT)
- Efficient measure to compute the matrix of DFT.