小波提取背景

用小波基估计背景信号

安装

不能直接安装：
pip install pywt
这是另外一个不同的库，如果不幸安装了，需要

pip uninstall pywt
pip uninstall PyWavelets

然后：

conda install PyWavelets
pip install PyWavelets

输出

添加背景信号

from pywt import wavedec,waverec
import numpy as np
import matplotlib.pyplot as plt
t = np.linspace(-100,100,50)
bg = 5
y = 50*np.exp(-t**2/(2*2^2)) +bg
plt.plot(t,y)

小波基估计背景

需要截去高频信息，然后用小波基估计出背景信号

iter = 10
x_filt = y
th = 1
for i in range(iter):
     #wavelet transform

    coeffs = wavedec(x_filt,'db6',level=5)
    cA = coeffs[0]
    cD =  []
    # 令高频信息为0，也就是令cDn全部设置为0
    for j in range(1,len(coeffs)):     
         a = np.squeeze( np.array([np.zeros([np.size(coeffs[j]),1 ])]) )
         cD.append(a)  
    cD.insert(0,cA)

    # inverse wavelet transform by only using low-freq components  
    x_new = waverec(cD, 'db6')
    if th > 0:
    # cut off values over current estimated background level.
        eps = np.sqrt(abs(x_filt))/2
        ind = y > (x_new + eps)
        x_filt[ind] = x_new[ind]+eps[ind]
        
        # re-estimate background 
        coeffs = wavedec(x_filt,'db6',level=5)
        cA = coeffs[0]
        cD =  []
        for j in range(1,len(coeffs)):     #排除cA5,cDn全部设置为0
             a = np.squeeze( np.array([np.zeros([np.size(coeffs[j]),1 ])]) )
             cD.append(a)  
        cD.insert(0,cA)

        x_new = waverec(cD, 'db6')

黄色的点表示估计出来的背景