How to find signal peaks with SciPy

Peak detection turns a sampled signal into the positions of meaningful local maxima, which helps identify pulses, beats, spikes, and repeated sensor responses. SciPy provides scipy.signal.find_peaks() for one-dimensional arrays, so the same pattern works for measured data, simulated signals, and small diagnostic samples.

find_peaks() returns the integer indexes of accepted peaks and a properties dictionary for filters that are requested. Values such as peak_heights, prominences, and widths appear only when the matching condition is supplied, which keeps a simple local-maximum search from producing extra arrays.

Choose thresholds in the same sample units as the array being searched. A distance value counts array positions, height compares the sample value at the peak, and prominence measures how far the peak rises above the surrounding baseline; replacing or removing NaN values before detection avoids misleading local comparisons.

1. Create a sample script that imports find_peaks() and defines the one-dimensional signal.

   peak_find.py

   import numpy as np
   from scipy.signal import find_peaks
    
   signal = np.array([0.0, 1.2, 0.1, 0.8, 0.2, 3.1, 0.4, 0.7, 0.2, 2.5, 0.1])
    
   peaks, properties = find_peaks(
       signal,
       height=1.0,
       distance=3,
       prominence=1.0,
       width=(None, None),
   )
    
   print("peak indexes:", peaks.tolist())
   print("peak values:", signal[peaks].round(2).tolist())
   print("prominences:", properties["prominences"].round(2).tolist())
   print("widths:", properties["widths"].round(2).tolist())
   print("expected match:", np.array_equal(peaks, np.array([1, 5, 9])))

   height filters by peak amplitude, distance requires spacing in samples, and prominence keeps peaks that stand above nearby baseline. width=(None, None) asks SciPy to calculate width properties without excluding peaks by width.

2. Run the script to verify the expected peak indexes.

   $ python3 peak_find.py
   peak indexes: [1, 5, 9]
   peak values: [1.2, 3.1, 2.5]
   prominences: [1.1, 3.0, 2.3]
   widths: [0.96, 1.07, 0.98]
   expected match: True

   The indexes point back into the original array. In this signal, indexes 1, 5, and 9 pass the height, spacing, and prominence conditions.

3. Replace the sample array with measured data and tune one condition at a time.

   Start with height or prominence, then add distance when neighboring detections must be separated by a minimum number of samples.

4. Remove the sample script after copying the pattern into your project.

   $ rm peak_find.py