gastropy.hampel_filter

Contents

gastropy.hampel_filter#

gastropy.hampel_filter(data, k=3, n_sigma=3.0)[source]#

Remove spike artifacts using the Hampel median identifier.

Slides a window of length 2k+1 across the signal. Any sample that deviates from the local median by more than n_sigma * 1.4826 * MAD is replaced by the local median.

The factor 1.4826 makes the scaled MAD a consistent estimator of the standard deviation under a Gaussian assumption (Davies & Gather, 1993).

Parameters:

  • data (array_like) – EGG signal(s). Accepts shape (n_samples,) or (n_channels, n_samples).

  • k (int, optional) – Half-window size: each sample is compared to its k neighbours on either side (window length = 2k+1). Default is 3.

  • n_sigma (float, optional) – Outlier threshold in scaled MAD units. Samples further than n_sigma from the local median are replaced. Default is 3.0.

Returns:

cleaned – Signal with spike outliers replaced by local medians. Same shape as input.

Return type:

np.ndarray

References

Davies, P. L., & Gather, U. (1993). The identification of multiple outliers. Journal of the American Statistical Association, 88, 782–792.

Dalmaijer, E. S. (2025). electrography v1.1.1. esdalmaijer/electrography

Examples

>>> import numpy as np
>>> from gastropy.signal import hampel_filter
>>> sig = np.sin(2 * np.pi * 0.05 * np.arange(0, 300, 0.1))
>>> sig[50] = 100.0  # inject a spike
>>> cleaned = hampel_filter(sig)
>>> np.abs(cleaned[50]) < 5.0  # spike removed
True
Contents