周期信号sin(wt): Fourier逆变换、变换

Universe-hello

于 2024-04-14 16:07:10 发布

阅读量251

点赞数 8

分类专栏：信号与傅立叶文章标签：信号处理傅立叶分析

本文链接：https://blog.csdn.net/weixin_51133173/article/details/137745688

版权

信号与傅立叶专栏收录该内容

15 篇文章 0 订阅

订阅专栏

文章介绍了如何使用Python中的numpy和matplotlib库实现Fourier变换，展示了一段关于正弦信号的频谱分析过程，包括计算频率、绘制频率-幅度图，并演示了如何过滤负频率并找出信号峰值。

摘要由CSDN通过智能技术生成

f=2, fs=200

$\boldsymbol{sin(\omega{t})=sin(2\pi{f}t)=sin(4\pi{t})}$

$\boldsymbol{t: 0\to{1s}}$

import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
_, ax = plt.subplots()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)

t0, t1 = 0,1
fs = 200
t = np.arange(t0,t1,1/fs)
f = 2
ft = np.sin(2*np.pi*f*t)

plt.scatter(t,ft,marker='.',s=1,c='b')
plt.savefig('f.jpg')

$\boldsymbol{f(t)=\frac{1}{t_1-t_0}\displaystyle\sum_{n=-\infty}^{n=\infty}\int_{t_0}^{t_1}f(t)e^{-i\omega_n{t}}dt\space e^{i\omega_n{t}}, (\omega_n=n\omega,n=...,-2,-1,0,1,2,...)}$

$\boldsymbol{F(\omega_n)=\int_{t_0}^{t_1}f(t)e^{-i\omega_n{t}}dt, (n=...,-2,-1,0,1,2,...)}$

$\boldsymbol{F^{-1}[F(\omega_n)](t)=\frac{1}{t_1-t_0}\displaystyle\sum_{n=-\infty}^{n=\infty}F(\omega_n) e^{i\omega_n{t}}, (n=...,-2,-1,0,1,2,...)}$

wn计算

$\boldsymbol{\omega_n=2\pi{f}n, (n=...,-2,-1,0,1,2,...)}$

>>> import numpy as np
>>> N = 3
>>> np.array(list(range(-N,N+1)))
array([-3, -2, -1,  0,  1,  2,  3])

>>> f = 2
>>> 2*np.pi*f*np.array(list(range(-N,N+1)))
array([-37.69911184, -25.13274123, -12.56637061,   0.        ,
        12.56637061,  25.13274123,  37.69911184])

Fourier逆变换

N=1

import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
_, ax = plt.subplots()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)

t0, t1 = 0,1
fs = 200
t = np.arange(t0,t1,1/fs)
f = 2
ft = np.sin(2*np.pi*f*t)
plt.scatter(t,ft,marker='.',s=1,c='b')

N = 1
W_n = 2*np.pi*f*np.array(list(range(-N,N+1)))
F_wn = lambda wn: sum(ft*np.exp(-1j*wn*t)*1/fs)
F_t = lambda ti: sum([F_wn(wn)*np.exp(1j*wn*ti) for wn in W_n])/(t1-t0)
Fourier_t = []
for i in tqdm(t):
    Fourier_t.append(F_t(i))
print('虚部的均值:',np.mean([i.imag for i in Fourier_t]))
Fourier_real = [i.real for i in Fourier_t]
plt.scatter(t,Fourier_real,marker='.',s=1,c='r')

plt.savefig('f.jpg')

N=2

N=10

Fourier变换

N=1

import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
_, ax = plt.subplots()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)

t0, t1 = 0,1
fs = 200
t = np.arange(t0,t1,1/fs)
f = 2
ft = np.sin(2*np.pi*f*t)

N = 1
W_n = 2*np.pi*f*np.array(list(range(-N,N+1)))
F_wn = lambda wn: sum(ft*np.exp(-1j*wn*t)*1/fs)
Fw = [F_wn(wn) for wn in W_n]
print(Fw)

[(-5.800481622797449e-18+0.5000000000000001j), (9.952975943416931e-17+0j), (-5.800481622797449e-18-0.5000000000000001j)]

是复数，用复数模长代替

Fw = [abs(F_wn(wn)) for wn in W_n]
print(Fw)

[0.5000000000000001, 9.952975943416931e-17, 0.5000000000000001]

(wn,fw)画图

import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
_, ax = plt.subplots()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)

t0, t1 = 0,1
fs = 200
t = np.arange(t0,t1,1/fs)
f = 2
ft = np.sin(2*np.pi*f*t)

N = 1
W_n = 2*np.pi*f*np.array(list(range(-N,N+1)))
F_wn = lambda wn: sum(ft*np.exp(-1j*wn*t)*1/fs)
Fw = [abs(F_wn(wn)) for wn in W_n]

plt.scatter(W_n,Fw,marker='*',s=30,c='r')
plt.savefig('f.jpg')

(频率,fw)画图

$\boldsymbol{\frac{\omega_n}{2\pi}=f_n}$

import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
_, ax = plt.subplots()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)

t0, t1 = 0,1
fs = 200
t = np.arange(t0,t1,1/fs)
f = 2
ft = np.sin(2*np.pi*f*t)

N = 1
W_n = 2*np.pi*f*np.array(list(range(-N,N+1)))
F_wn = lambda wn: sum(ft*np.exp(-1j*wn*t)*1/fs)
Fw = [abs(F_wn(wn)) for wn in W_n]

freq_n = [wn/(2*np.pi) for wn in W_n]
plt.scatter(freq_n,Fw,marker='*',s=30,c='r')
plt.savefig('f.jpg')

n可以取负数，是因为Fourier级数-复数形式，有如下关系

$\boldsymbol{\displaystyle\sum_{n=1}^\infty\int_{\frac{-T}{2}}^{\frac{T}{2}}f_T(t)e^{in\omega{t}}dt\space e^{-in\omega{t}}\space= \displaystyle\sum_{-\infty}^{n=-1}\space\int_{\frac{-T}{2}}^{\frac{T}{2}}f_T(t)e^{-in\omega{t}}dt\space e^{in\omega{t}}}$

为了公式的简洁性，把n从1到inf推到了-inf~inf

在本文章中，我们想从实际意义的角度看信号跟频率的关系，

可以只看Fourier变换后: w或f为正的那一半，但是Fourier变换公式计算的过程中正负频率都要算

过滤n<0

import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
_, ax = plt.subplots()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)

t0, t1 = 0,1
fs = 200
t = np.arange(t0,t1,1/fs)
f = 2
ft = np.sin(2*np.pi*f*t)

N = 1
W_n = 2*np.pi*f*np.array(list(range(-N,N+1)))
F_wn = lambda wn: sum(ft*np.exp(-1j*wn*t)*1/fs)
Fw = [abs(F_wn(wn)) for wn in W_n[W_n>=0]]

freq_n = [wn/(2*np.pi) for wn in W_n[W_n>=0]]
plt.scatter(freq_n,Fw,marker='*',s=30,c='r')
index = np.argmax(Fw)
plt.text(freq_n[index]+0.1,Fw[index],f'freq={freq_n[index]}')
plt.savefig('f.jpg')

N=2

import numpy as np
import matplotlib.pyplot as plt
from tqdm import tqdm
_, ax = plt.subplots()
ax.spines['right'].set_visible(False)
ax.spines['top'].set_visible(False)

t0, t1 = 0,1
fs = 200
t = np.arange(t0,t1,1/fs)
f = 2
ft = np.sin(2*np.pi*f*t)

N = 2
W_n = 2*np.pi*f*np.array(list(range(-N,N+1)))
F_wn = lambda wn: sum(ft*np.exp(-1j*wn*t)*1/fs)
Fw = [abs(F_wn(wn)) for wn in W_n[W_n>=0]]

freq_n = [wn/(2*np.pi) for wn in W_n[W_n>=0]]
plt.scatter(freq_n,Fw,marker='*',s=30,c='r')
index = np.argmax(Fw)
plt.text(freq_n[index]+0.1,Fw[index],f'freq={freq_n[index]}')
plt.savefig('f.jpg')

N=10

Universe-hello

关注

8
点赞
踩
1

收藏

觉得还不错? 一键收藏
0
评论
周期信号sin(wt): Fourier逆变换、变换

可以只看Fourier变换后: w或f为正的那一半，但是Fourier变换公式计算的过程中正负频率都要算。为了公式的简洁性，把n从1到inf推到了-inf~inf。在本文章中，我们想从实际意义的角度看信号跟频率的关系，，是因为Fourier级数-复数形式，有如下关系。
复制链接

扫一扫

专栏目录