我有一个重复信号,每个周期重复大约每秒一次,但是每个周期的持续时间和内容在某些参数内略有不同。每秒信号数据都有一千个x,y坐标。每个周期内的一小部分但很重要的数据部分已损坏,我想用向上的抛物线替换每个损坏的部分。
对于需要用抛物线替换的每个数据段,我有三个点的x,y坐标。顶点/最小值是这些点之一。另外两个点是向上U形的左右顶部,即抛物线。换句话说,左上角是此函数域中最低x值的x,y坐标对,而右上角是此函数域中最高x值的x,y坐标对。左上顶和右上顶点的y坐标彼此相等,是数据段中两个最高的y值。
如何编写代码来绘制这个面向上的抛物线中的剩余数据点?请记住,对于每分钟数据,此函数需要调用60或70次,并且形状为/抛物线的公式在每次调用此函数时都需要更改,以便考虑每个结果抛物线中这三对x,y坐标之间的不同关系。
def ReplaceCorruptedDataWithParabola(Xarray, Yarray, LeftTopX, LeftTopY
, LeftTopIndex, MinX, MinY, MinIndex
, RightTopX, RightTopY, RightTopIndex):
# Step One: Derive the formula for the upward-facing parabola using
# the following data from the three points:
LeftTopX,LeftTopY,LeftTopIndex
MinX,MinY,MinIndex
RightTopX,RightTopY,RightTopIndex
# Step Two: Use the formula derived in step one to plot the parabola in
# the places where the corrupted data used to reside:
for n in Xarray[LeftTopX:RightTopX]:
Yarray[n]=[_**The formula goes here**_]
return Yarray
注意:Xarray和Yarray是每个单列向量,每个索引都有数据,将两个数组链接为x,y坐标集。它们都是numpy数组。 Xarray包含时间信息但不会改变,但是Yarray包含信号数据,包括将被需要由此函数计算的抛物线数据替换的损坏段。
答案 0 :(得分:7)
所以,据我所知,你有3个点你想要抛物线。
通常情况下,使用numpy.polyfit最简单,但如果你真的担心速度,并且你恰好适合三个点,那么使用最小二乘拟合没有意义。
相反,我们有一个均匀确定的系统(将抛物线拟合到3 x,y点),我们可以得到一个简单线性代数的精确解。
所以,总而言之,你可能会做这样的事情(大部分是绘制数据):
import numpy as np
import matplotlib.pyplot as plt
def main():
# Generate some random data
x = np.linspace(0, 10, 100)
y = np.cumsum(np.random.random(100) - 0.5)
# Just selecting these arbitrarly
left_idx, right_idx = 20, 50
# Using the mininum y-value within the arbitrary range
min_idx = np.argmin(y[left_idx:right_idx]) + left_idx
# Replace the data within the range with a fitted parabola
new_y = replace_data(x, y, left_idx, right_idx, min_idx)
# Plot the data
fig = plt.figure()
indicies = [left_idx, min_idx, right_idx]
ax1 = fig.add_subplot(2, 1, 1)
ax1.axvspan(x[left_idx], x[right_idx], facecolor='red', alpha=0.5)
ax1.plot(x, y)
ax1.plot(x[indicies], y[indicies], 'ro')
ax2 = fig.add_subplot(2, 1, 2)
ax2.axvspan(x[left_idx], x[right_idx], facecolor='red', alpha=0.5)
ax2.plot(x,new_y)
ax2.plot(x[indicies], y[indicies], 'ro')
plt.show()
def fit_parabola(x, y):
"""Fits the equation "y = ax^2 + bx + c" given exactly 3 points as two
lists or arrays of x & y coordinates"""
A = np.zeros((3,3), dtype=np.float)
A[:,0] = x**2
A[:,1] = x
A[:,2] = 1
a, b, c = np.linalg.solve(A, y)
return a, b, c
def replace_data(x, y, left_idx, right_idx, min_idx):
"""Replace the section of "y" between the indicies "left_idx" and
"right_idx" with a parabola fitted to the three x,y points represented
by "left_idx", "min_idx", and "right_idx"."""
x_fit = x[[left_idx, min_idx, right_idx]]
y_fit = y[[left_idx, min_idx, right_idx]]
a, b, c = fit_parabola(x_fit, y_fit)
new_x = x[left_idx:right_idx]
new_y = a * new_x**2 + b * new_x + c
y = y.copy() # Remove this if you want to modify y in-place
y[left_idx:right_idx] = new_y
return y
if __name__ == '__main__':
main()
希望有所帮助...