用移动窗口的正态分布替换 nans

Question

我需要用numpy中的局部正态分布替换一维数组的NaN。我选择一个窗口，计算该窗口的均值和标准差，然后使用正态分布替换NaN，而其余信号保持不变。

import numpy as np

def replace_nan(signal, window = 5):

    """

    calculate moving average and std of signal window without nan    

    replaces nan values with normal distribution (mean, std)    

    """

# add padding in case signal starts/ends with nan

    signal = np.pad(signal, (window, window), 'mean', stat_length = 2*window)    

    for k in range(window,len(signal)-window):        

        mean = np.nanmean(signal[k-window:k+window])  # window average 

        std = np.nanstd(signal[k-window:k+window]) # window std without nan 

        ind = np.where(np.isnan(signal[k-window:k+window]))[0]    

        print (ind)   

        signal[ind]= np.random.normal(mean, std)

    signal = signal[window:len(signal)-window] #remove padding

    return signal

#tester 

signal = np.array([0.71034849, 0.17730998, 0.77577915, 0.38308111, 

0.24278947, np.nan, np.nan, 0.68694097, 0.6684736 , 0.47310845, 0.22210945, 

0.1189111, np.nan, np.nan, np.nan, 0.5573841 , 0.57531205, 0.74131346, 

0.29088101, 0.5573841 , 0.57531205, 0.74131346, np.nan, np.nan, np.nan, 

np.nan, 0.49534304, 0.18370482, 0.06089498, 0.22210945, 0.1189111])        

signal = replace_nan(signal, 5)

print(signal)

我将 nans 替换为正态分布 np.random.normal() ，并为大小为 5 的移动窗口计算均值和标准差。当我选择信号窗口的那些 nans 来替换它们时，出现了问题。这应该很容易，我只是python的完整入门者。

Answer 1

我还没有测试过这些数字是否准确，但是我认为这会起作用：import numpy as npdef replace_nan(signal, window = 5):    """    calculate moving average and std of signal window without nan    replaces nan values with normal distribution (mean, std)    """    # add padding in case signal starts/ends with nan    signal = np.pad(signal, (window, window), 'mean', stat_length = 2*window)    for k in range(window, len(signal) - window + 1):        mean = np.nanmean(signal[k-window:k+window])  # window average        std = np.nanstd(signal[k-window:k+window]) # window std without nan        if np.isnan(signal[k]):            signal[k] = np.random.normal(mean, std)    signal = signal[window:len(signal)-window] #remove padding    return signal#testersignal = np.array(    [        0.71034849, 0.17730998, 0.77577915, 0.38308111, 0.24278947, np.nan,        np.nan, 0.68694097, 0.6684736 , 0.47310845, 0.22210945, 0.1189111,        np.nan, np.nan, np.nan, 0.5573841 , 0.57531205, 0.74131346,        0.29088101, 0.5573841 , 0.57531205, 0.74131346, np.nan, np.nan,        np.nan, np.nan, 0.49534304, 0.18370482, 0.06089498, 0.22210945,        0.1189111    ])print("Before:")print(signal)signal = replace_nan(signal, 5)print("\nAfter:")print(signal)这给出了：Before:[ 0.71034849  0.17730998  0.77577915  0.38308111  0.24278947         nan         nan  0.68694097  0.6684736   0.47310845  0.22210945  0.1189111         nan         nan         nan  0.5573841   0.57531205  0.74131346  0.29088101  0.5573841   0.57531205  0.74131346         nan         nan         nan         nan  0.49534304  0.18370482  0.06089498  0.22210945  0.1189111 ]After:[ 0.71034849  0.17730998  0.77577915  0.38308111  0.24278947  0.35960417  0.508657    0.68694097  0.6684736   0.47310845  0.22210945  0.1189111  0.50282732  0.34906067  0.31206557  0.5573841   0.57531205  0.74131346  0.29088101  0.5573841   0.57531205  0.74131346  0.80133879  0.63122315  0.49236281  0.35630875  0.49534304  0.18370482  0.06089498  0.22210945  0.1189111 ]