Find Start and End point of a Curve in Python

Question

I have a signal with 5 curves and my goal is to find the start, peak, and end point of each curve.

To find the peaks, I am using find_peaks function from scipy.signal. The following screenshot shows signal and the 5 peaks:

In order to find the start point of each of these curves, I have written a function that checks for the percentage difference between every pair of signal values. If the difference is less than or equal to 1%, then I mark it as the start point.

def find_start(peak, leftLimit, signal):
    # Go in left direction from the peak
    # Start at the current curve's peak, and go until the previous curve's peak (or until zero in case of very first curve)
    for i in range(peak, leftLimit, -1):
        diff = signal[i] - signal[i-1]
        perc_change = float(diff)/signal[i] * 100
        perc_change = round(abs(perc_change), 3)
        if perc_change <= 1.0:
            return i
    return -1

I have a similar function to find the end point of each curve.

Now, although these two functions work in general, they fail if the signal is quite distorted and is not constant (straight horizontal line) outside the 5 curves. The following screenshot shows one such case:

Is there a better way (or any Python Library function) that can help locate the start and end points for each curve?

Answer 1

I would try folding your signal with a defined pulse (rectangular or triangle) and shift this triangle over your signal.

It is basically described here: https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.correlate.html

In the resulting correlation result you could apply a simple thresholding to get the indices where it rises/dropes over/below the threshold.