注:英文引文机翻,未校。
如有内容异常,请看原文。
Euler's Formula and Fourier Transform
Posted byczxttkl October 7, 2018
Euler's formula states that e i x = cos x + i sin x e^{ix} =\cos{x}+ i \sin{x} eix=cosx+isinx. When x = π x = \pi x=π, the formula becomes e π = − 1 e^{\pi} = -1 eπ=−1 known as Euler's identity.
欧拉公式表明 e i x = cos x + i sin x e^{ix} = \cos{x} + i \sin{x} eix=cosx+isinx。当 x = π x = \pi x=π 时,公式变为 e π = − 1 e^{\pi} = -1 eπ=−1,称为欧拉恒等式。
An easy derivation of Euler's formula is given in [3] and [5]. According to Maclaurin series (a special case of taylor expansion f ( x ) = f ( a ) + f ′ ( a ) ( x − a ) + f " ( a ) 2 ! ( x − a ) 2 + ⋯ f(x)=f(a)+f'(a)(x-a)+\frac{f"(a)}{2!}(x-a)^2+\cdots f(x)=f(a)+f′(a)(x−a)+2!f"(a)(x−a)2+⋯ when a = 0 a=0 a=0),
一个简单的欧拉公式推导见于 [3] 和 [5]。根据麦克劳林级数(泰勒展开的一个特例 f ( x ) = f ( a ) + f ′ ( a ) ( x − a ) + f ′ ′ ( a ) 2 ! ( x − a ) 2 + ⋯ f(x) = f(a) + f'(a)(x-a) + \frac{f''(a)}{2!}(x-a)^2 + \cdots f(x)=f(a)+f′(a)(x−a)+2!f′′(a)(x−a)2+⋯ 当 a = 0 a = 0 a=0 时),
e x = 1 + x + x 2 2 ! + x 3 3 ! + x 4 4 ! + ⋯ = 2 e^x=1+x+\frac{x^2}{2!}+\frac{x^3}{3!}+\frac{x^4}{4!}+\cdots =2 ex=1+x+2!x2+3!x3+4!x4+⋯=2
Therefore, replacing x x x with i x ix ix, we have
因此,将 x x x 替换为 i x ix ix,我们得到
e i x = 1 + i x − x 2 2 ! − x 3 3 ! + x 4 4 ! + x 5 5 ! − ⋯ = 2 e^{ix}=1+ix-\frac{x^2}{2!}-\frac{x^3}{3!}+\frac{x^4}{4!}+\frac{x^5}{5!}-\cdots =2 eix=1+ix−2!x2−3!x3+4!x4+5!x5−⋯=2
By Maclaurin series, we also have
根据麦克劳林级数,我们还有
cos x = 1 − x 2 2 ! + x 4 4 ! − x 6 6 ! + ⋯ sin x = x − x 3 3 ! + x 5 5 ! − ⋯ = 2 \cos{x}=1-\frac{x^2}{2!}+\frac{x^4}{4!}-\frac{x^6}{6!} + \cdots \newline \sin{x}=x -\frac{x^3}{3!}+\frac{x^5}{5!}-\cdots =2 cosx=1−2!x2+4!x4−6!x6+⋯sinx=x−3!x3+5!x5−⋯=2
Therefore, we can rewrite e i x e^{ix} eix as e i x = cos x + i sin x e^{ix}=\cos{x}+i\sin{x} eix=cosx+isinx
因此,我们可以将 e i x e^{ix} eix 重写为 e i x = cos x + i sin x e^{ix} = \cos{x} + i \sin{x} eix=cosx+isinx。
Intuitive understanding of e i x = cos x + i sin x e^{ix}=\cos{x}+i\sin{x} eix=cosx+isinx is illustrated in [1] together with its own video [4], as well as in 3Blue1Brown's two videos [6][7].
e i x = cos x + i sin x e^{ix} = \cos{x} + i \sin{x} eix=cosx+isinx 的直观理解在 [1] 中有说明,连同其自己的视频 [4],以及在 3Blue1Brown 的两个视频 [6][7] 中。
First, from a conventional view of coordinate system, e i x = cos x + i sin x e^{ix}=\cos{x}+i\sin{x} eix=cosx+isinx means a point with x coordinate cos x \cos{x} cosx and y coordinate sin x \sin{x} sinx on a unit circle (centered at the origin with radius 1) in the complex plane.
首先,从常规的坐标系视角来看, e i x = cos x + i sin x e^{ix} = \cos{x} + i \sin{x} eix=cosx+isinx 表示复平面上单位圆(以原点为中心,半径为1)上的一个点,其 x 坐标为 cos x \cos{x} cosx,y 坐标为 sin x \sin{x} sinx。
Another view is that e i x e^{ix} eix describes the point that moves distance x x x from (1,0) along the circumference of the unit circle.
另一种观点是 e i x e^{ix} eix 描述了一个点,它从 (1,0) 沿单位圆周移动距离 x x x。
The last view, which is similar to the second view, is that e i x e^{ix} eix specifies the point such that the degree between the x axis and the line connects that point to the origin is radiant x x x.
最后一个观点与第二个观点类似, e i x e^{ix} eix 指定了一个点,使得x轴与连接该点到原点的线之间的夹角是 x x x 弧度。
For example, e i ⋅ 1 e^{i \cdot 1} ei⋅1 viewed in x&y coordinates:
例如, e i ⋅ 1 e^{i \cdot 1} ei⋅1 在 x&y 坐标中查看:

or viewed from the moving perspective:
或从移动的视角查看:

or viewed from the radiant perspective:
或从辐射的视角查看:

As pointed out by [1], we can think of normal exponential as stretching an axis such that number 1 is stretched to a point denoting the exponential result. For example, 2 3 2^3 23 means 1 is stretched to 8, or 1 is stretched to 2 first, then that point is stretched to 4, then finally that point is stretched to 8. However, complex exponential e i x e^{ix} eix means to rotate from the point (1,0) at a constant rotation speed x x x. Therefore, Euler's identity can be interpreted as starting from point (1,0), a point moves half of the circle (radiant π \pi π) and ends up at (-1, 0), therefore e π = − 1 e^{\pi}=-1 eπ=−1. A good diagram summarizing this intuition is as below:
正如 [1] 所指出的,我们可以将普通的指数视为拉伸一个轴,使得数字1被拉伸到表示指数结果的点。例如, 2 3 2^3 23 意味着 1 被拉伸到 8,或者 1 先被拉伸到 2,然后那个点被拉伸到 4,最后那个点被拉伸到 8。然而,复数指数 e i x e^{ix} eix 意味着从点 (1,0) 开始以 恒定 的旋转速度 x x x 旋转 。因此,欧拉恒等式可以被解释为从点 (1,0) 开始,一个点移动了半个圆( π \pi π 弧度)并最终到达 (-1, 0),因此 e π = − 1 e^{\pi} = -1 eπ=−1。一个很好地总结这种直观理解的示意图如下:

Why understanding Euler's formula as a point rotating on a unit circle helps? One example in which Euler's formula is useful is Fourier transform. The goal of Fourier transform is to turning signals measured in time space into analyzable, frequency-based spectrum. Fourier transform is intuitively illustrated in BetterExplained [8, 3Blue1Brown's [9], and Math StackExchange [16]. Below is notes taken based on [9].
为什么理解欧拉公式作为单位圆上的一个旋转点是有帮助的?欧拉公式有用的一个例子是傅里叶变换。傅里叶变换的目标是将时间空间中测量的信号转换成可分析的、基于频率的频谱。傅里叶变换在 BetterExplained [8, 3Blue1Brown's] [9] 和 Math StackExchange 16 16 16 中直观地说明了。以下是根据 [9] 记录的笔记。
A concrete example to apply Fourier transform is to analyze a signal which mixes two different signals, each of a constant frequency. In the example, suppose we can only measure the blended signal (green), while we want to tell posthumously that it actually composes of the pink signal (for example, 2Hz) and the yellow signal (for example, 3Hz).
应用傅里叶变换的一个具体例子是分析混合了两个不同信号的信号,每个信号都有恒定的频率。在例子中,假设我们只能测量混合信号(绿色),而我们想要事后说明它实际上由粉色信号(例如,2Hz)和黄色信号(例如,3Hz)组成。

The magic idea of Fourier transform is that if you transform the signal measured in the Pressure vs. Time space into a rotating motion in a 2D plane, it will looks like the diagram below:
傅里叶变换的神奇之处在于,如果你将压力与时间空间中测量的信号转换成二维平面中的旋转运动,它将看起来像下面的图表:

Depending on the frequency of rotating motion, the diagram on the 2D plane may be different.
根据旋转运动的频率,二维平面上的图表可能会有所不同。

An important observation is that if you look at the center of mass of the diagram based on any rotating frequency, you will see that the center of mass is farthest away from the origin when the frequency of rotating motion matches the frequency of either individual signal (2Hz or 3Hz).
一个重要的观察是,如果你查看基于任何旋转频率的图表的质心,你会看到当旋转运动的频率与任一单独信号的频率(2Hz或3Hz)匹配时,质心距离原点最远。

Therefore, to recover original signals, we need to try out every possible frequency of rotating motion and just obtain the location of the center of mass of the diagram for each frequency of rotating motion, then we would know what are the frequencies of the original individual signals.
因此,为了恢复原始信号,我们需要尝试每一种可能的旋转运动频率,并仅获得每种旋转运动频率的图表质心的位置,然后我们就会知道原始单独信号的频率是什么。
The center of mass can be obtained by sampling and averaging of the data points on the diagram, while the diagram is mapped onto a 2D complex plane:
质心可以通过对图表上的数据点进行采样和平均来获得,而图表被映射到二维复平面上:

When the number of samples goes to infinite, it becomes integral.
当样本数量趋向于无穷大时,它就变成了积分。

Fourier transform, following this idea, is only different in that the integral is from negative infinity to positive infinity, and there is no final average over time ( 1 t 2 − t 1 \frac{1}{t_2-t_1} t2−t11 removed).
傅里叶变换,遵循这个想法,唯一的不同在于积分是从负无穷到正无穷,并且没有最终的时间平均( 1 t 2 − t 1 \frac{1}{t_2-t_1} t2−t11 被移除)。

Here is an example [15] where Fourier transform is applied on f ( t ) = cos ( 2 π s t ) f(t)=\cos(2\pi st) f(t)=cos(2πst). The result is two Dirac Delta functions on the frequency domain.
这里有一个例子 [15],傅里叶变换应用于 f ( t ) = cos ( 2 π s t ) f(t) = \cos(2\pi st) f(t)=cos(2πst)。结果是频率域上的两个狄拉克δ函数。

The transform process is based on trigonometry. In the second to the last equation, when u = ± s u = \pm s u=±s,
变换过程基于三角学。在倒数第二个方程中,当 u = ± s u = \pm s u=±s 时,
∫ − ∞ ∞ cos ( 2 π s t ) cos ( 2 π u t ) d t = ∫ − ∞ ∞ cos 2 ( 2 π s t ) d t = ∫ − ∞ ∞ 1 2 d t + 1 2 ∫ − ∞ ∞ cos ( 4 π s t ) d t = 1 2 ⋅ ∞ + 1 2 ∫ − ∞ ∞ cos ( 4 π s t ) d t ( s = 2 ) \begin{align*} \int_{-\infty}^{\infty} \cos(2\pi st) \cos(2\pi ut) dt &=\int_{-\infty}^{\infty} \cos^2(2\pi st) dt\\ &= \int_{-\infty}^{\infty} \frac{1}{2} dt + \frac{1}{2} \int_{-\infty}^{\infty} \cos(4\pi st) dt\\ &= \frac{1}{2} \cdot \infty + \frac{1}{2} \int_{-\infty}^{\infty} \cos(4\pi st) dt \quad (s = 2) \end{align*} ∫−∞∞cos(2πst)cos(2πut)dt=∫−∞∞cos2(2πst)dt=∫−∞∞21dt+21∫−∞∞cos(4πst)dt=21⋅∞+21∫−∞∞cos(4πst)dt(s=2)
According to [10], although ∫ − ∞ ∞ cos ( 4 π s t ) d t ( s = 2 ) \int_{-\infty}^{\infty} \cos(4\pi st) dt \quad (s = 2) ∫−∞∞cos(4πst)dt(s=2) does not converge, I have seen many places [15, 18] treating ∫ − ∞ ∞ cos ( 4 π s t ) d t = 0 ( s = 2 ) \int_{-\infty}^{\infty}\cos(4\pi st) dt = 0 \quad (s = 2) ∫−∞∞cos(4πst)dt=0(s=2). I don't understand this, but I could think it in an intuitive way: although ∫ − ∞ ∞ cos ( 4 π s t ) d t ( s = 2 ) \int_{-\infty}^{\infty}\cos(4\pi st) dt \quad (s = 2) ∫−∞∞cos(4πst)dt(s=2) does not converge, its values is confined in a limited range ( ≤ 2 \leq 2 ≤2). Compared to the infinity we obtain from ∫ − ∞ ∞ 1 2 d t ( s = 2 ) \int_{-\infty}^{\infty} \frac{1}{2} dt \quad (s = 2) ∫−∞∞21dt(s=2), the value of ∫ − ∞ ∞ cos ( 4 π s t ) d t ( s = 2 ) \int_{-\infty}^{\infty}\cos(4\pi st) dt \quad (s = 2) ∫−∞∞cos(4πst)dt(s=2) is infinitely close to zero.
根据 [10],尽管 ∫ − ∞ ∞ cos ( 4 π s t ) d t ( s = 2 ) \int_{-\infty}^{\infty} \cos(4\pi st) dt \quad (s = 2) ∫−∞∞cos(4πst)dt(s=2) 不收敛,但我在很多地方 [15, 18] 看到将 ∫ − ∞ ∞ cos ( 4 π s t ) d t = 0 ( s = 2 ) \int_{-\infty}^{\infty}\cos(4\pi st) dt = 0 \quad (s = 2) ∫−∞∞cos(4πst)dt=0(s=2) 来处理的情况。我不理解这一点,但我可以用一种直观的方式来思考:尽管 ∫ − ∞ ∞ cos ( 4 π s t ) d t ( s = 2 ) \int_{-\infty}^{\infty}\cos(4\pi st) dt \quad (s = 2) ∫−∞∞cos(4πst)dt(s=2) 不收敛,但其值被限制在一个有限的范围内( ≤ 2 \leq 2 ≤2 )。与我们从 ∫ − ∞ ∞ 1 2 d t ( s = 2 ) \int_{-\infty}^{\infty} \frac{1}{2} dt \quad (s = 2) ∫−∞∞21dt(s=2) 得到的无穷大相比, ∫ − ∞ ∞ cos ( 4 π s t ) d t ( s = 2 ) \int_{-\infty}^{\infty}\cos(4\pi st) dt \quad (s = 2) ∫−∞∞cos(4πst)dt(s=2) 的值无限接近于零。
Another idea is to solve the integration based on Euler's formula [17][20]. Based on Euler's formula, we can get cos ( x ) = 1 2 ( e i x + e − i x ) ( s = 2 ) \cos(x)=\frac{1}{2}(e^{ix} + e^{-ix}) \quad (s = 2) cos(x)=21(eix+e−ix)(s=2). Also, Dirac Delta function is defined as (its proof, which I don't fully understand, can be found on [11, 12, 13]):
另一个思路是基于欧拉公式 [17][20] 来求解积分。基于欧拉公式,我们可以得到 cos ( x ) = 1 2 ( e i x + e − i x ) ( s = 2 ) \cos(x)=\frac{1}{2}(e^{ix} + e^{-ix}) \quad (s = 2) cos(x)=21(eix+e−ix)(s=2) 。此外,狄拉克δ函数定义如下(其证明我不完全理解,可以在 [11, 12, 13] 中找到):
δ ( x ) = 1 2 π ∫ − ∞ ∞ e − j x t d t ( s = 2 ) \delta(x)= \frac{1}{2\pi} \int_{-\infty}^{\infty} e^{-jxt}dt \quad (s = 2) δ(x)=2π1∫−∞∞e−jxtdt(s=2)
Or equivalently,
或者等价地,
δ ( x ) = ∫ − ∞ ∞ e − j 2 π x t d t ( s = 2 ) \delta(x) =\int_{-\infty}^{\infty} e^{-j2\pi xt}dt \quad (s = 2) δ(x)=∫−∞∞e−j2πxtdt(s=2)
Therefore, we can finally get (based on [17] [20], with a little difference because they transform cos ( ω 0 t ) \cos(\omega_0 t) cos(ω0t) rather than cos ( 2 π s t ) \cos(2\pi st) cos(2πst)):
因此,我们最终可以得到(基于 [17] [20],有一点不同,因为他们变换的是 cos ( ω 0 t ) \cos(\omega_0 t) cos(ω0t) 而不是 cos ( 2 π s t ) \cos(2\pi st) cos(2πst)):

According to the linear property of Fourier transform 19 19 19, if a blended signal is the sum of several signals (for example, cos ( 2 π s t ) + cos ( 3 π s t ) \cos(2\pi st)+\cos (3\pi st) cos(2πst)+cos(3πst), then the resultant Fourier transform is the sum of several Dirac functions. That's how Fourier transform extracts individual signals from the mixed one! And the connection between Euler's formula and Fourier transform is that the integration required by Fourier transform to obtain the center of mass of the rotation diagram, if connected to Euler's formula, can be calculated very easily.
根据傅里叶变换的线性性质 19 19 19,如果混合信号是几个信号的总和(例如, cos ( 2 π s t ) + cos ( 3 π s t ) \cos(2\pi st) + \cos(3\pi st) cos(2πst)+cos(3πst)),那么得到的傅里叶变换是几个狄拉克函数的总和。这就是傅里叶变换如何从混合信号中提取单个信号的!欧拉公式和傅里叶变换之间的联系在于,傅里叶变换为了获得旋转图表的质心所需的积分,如果与欧拉公式联系起来,可以非常容易地计算。
14 14 14 is another illustrative video showing that any signal we can measure can be actually seen as a combination of an infinite number of sine waves. For repeating waves in time space, its Fourier transform may shown as discrete frequency spectrum (like several Dirac deltas). However, non-repeating waves in time space may result to continuous frequency spectrum after Fourier transform. This is interesting to know but I will not explore further within this article.
14 14 14 是另一个说明性视频,显示任何我们可以测量的信号实际上可以看作是无限数量的正弦波的组合。对于时间空间中的重复波,其傅里叶变换可能显示为离散的频率谱(像几个狄拉克δ)。然而,时间空间中的非重复波在傅里叶变换后可能导致连续的频率谱。这是一件有趣的事情,但我将不在本文中进一步探讨。

References
1\]