PeakFinder Classes#

Base Class#

Moving-average Class#

class SlidingWindowPeakLocator(window_size, max_number_of_peaks=5, padding_value=-1, window_step=-1, compute_width=False, compute_area=False, threshold=0.5)[source]#

Bases: BasePeakLocator

SlidingWindowPeakLocator(window_size, window_step=-1, max_number_of_peaks=5, padding_value=-1,

compute_width=False, compute_area=False, threshold=0.5)

A sliding-window-based peak detection utility for 1D signals. This class segments the input signal into fixed-size windows (which can be overlapping if window_step is less than window_size) and identifies the local maximum in each window. Optionally, it computes additional metrics for each peak: - Width: the number of contiguous samples above a specified fraction of the peak’s height. - Area: the sum of signal values under the peak within its window.

The results are returned as a dictionary containing fixed-length arrays. If fewer peaks are detected than max_number_of_peaks, the arrays are padded with padding_value (for indices) or NaN (for numeric values).

Parameters:

  • window_size (int) – The size of the sliding window used for local peak detection.

  • window_step (int, optional) – The step size between consecutive windows. If not provided or set to -1, defaults to window_size (i.e., non-overlapping windows). To create overlapping windows, specify a value less than window_size.

  • max_number_of_peaks (int, optional) – The maximum number of peaks to report. If fewer peaks are detected, the results are padded. Default is 5.

  • padding_value (int, optional) – The value used to pad the output array for indices when fewer than max_number_of_peaks peaks are found. Default is -1.

  • compute_width (bool, optional) – If True, compute and return the width of each detected peak (in samples). Default is False.

  • compute_area (bool, optional) – If True, compute and return the area (sum of signal values) under each detected peak. Default is False.

  • threshold (float, optional) – The fraction of the peak’s height used to determine the boundaries for width and area calculations. For example, a threshold of 0.5 uses 50% of the peak height as the cutoff. Default is 0.5.

Returns:

A dictionary containing the following keys:

  • ”Index”: A fixed-length array of detected peak indices, padded as necessary.

  • ”Height”: A fixed-length array of the corresponding peak values.

  • ”Width”: (optional) A fixed-length array of computed peak widths (if compute_width is True).

  • ”Area”: (optional) A fixed-length array of computed peak areas (if compute_area is True).

Return type:

dict

Examples

>>> import numpy as np
>>> # Example signal
>>> data = np.array([1, 3, 2, 5, 4, 2, 8, 1], dtype=float)
>>> # Create a peak locator without additional metrics (non-overlapping windows).
>>> locator = SlidingWindowPeakLocator(window_size=2, max_number_of_peaks=4, padding_value=-1)
>>> result = locator(data)
>>> print("Indices:", result["Index"])
[1, 3, 4, 6]
>>> print("Heights:", result["Height"])
[3.0, 5.0, 4.0, 8.0]
>>> # Create a peak locator with overlapping windows and with width and area computations.
>>> locator_metrics = SlidingWindowPeakLocator(window_size=4, window_step=2, max_number_of_peaks=4,
...                                             padding_value=-1, compute_width=True, compute_area=True, threshold=0.5)
>>> result = locator_metrics(data)
>>> print("Peak indices:", result["Index"])
>>> print("Heights:", result["Height"])
>>> print("Widths:", result["Width"])
>>> print("Areas:", result["Area"])

Derivative based Class#

class DerivativePeakLocator(max_number_of_peaks=5, padding_value=-1, compute_width=False, compute_area=False, threshold=0.5)[source]#

Bases: BasePeakLocator

A peak detection utility based on the zero-crossing of the first derivative.

This class detects peaks by computing the first and second derivatives of each row of a 2D array. A peak is defined where the first derivative changes sign from positive to negative and the second derivative is negative. Optionally, for each detected peak, the width (number of samples above a threshold) and area (sum of values over the peak region) can be computed using a simple threshold-based approach.

Parameters:

  • max_number_of_peaks (int, optional) – Maximum number of peaks to return per row. If fewer peaks are detected, missing values are padded with padding_value. Default is 5.

  • padding_value (object, optional) – Value to pad missing peaks in output. Default is -1.

  • compute_width (bool, optional) – If True, compute the width (in samples) of each detected peak. Default is False.

  • compute_area (bool, optional) – If True, compute the area (sum of values) under each detected peak. Default is False.

  • threshold (float, optional) – Fraction of the peak height used to determine the boundaries for width and area calculations. Default is 0.5.

Examples

>>> import numpy as np
>>> # Without additional metrics:
>>> peak_locator = DerivativePeakLocator(max_number_of_peaks=3)
>>> data = np.array([[0, 100, 500, 200, 50, 1000],
...                  [200, 400, 100, 900, 500, 700]])
>>> result = peak_locator(data)
>>> print(result["peak_index"])
[[2. 5. nan]
 [1. 3. 5.]]

>>> # With width and area computations enabled:
>>> peak_locator_metrics = DerivativePeakLocator(max_number_of_peaks=3, compute_width=True, compute_area=True, threshold=0.5)
>>> result = peak_locator_metrics(data)
>>> print("Peak indices:\n", result["peak_index"])
>>> print("Widths:\n", result["width"])
>>> print("Areas:\n", result["area"])
On this page
Show Source