mne_denoise.dss.segmentation.CovarianceSegmenter

class mne_denoise.dss.segmentation.CovarianceSegmenter(sfreq: float, min_chunk_len: float = 30.0, cov_win_len: float = 1.0, bandpass: tuple[float, float] | None = None, prominence: float = 0.5)

Segment data where windowed covariance changes.

Parameters:

sfreqfloat

Sampling frequency in Hz.

min_chunk_lenfloat, default=30.0

Minimum segment length in seconds.

cov_win_lenfloat, default=1.0

Covariance-window length in seconds.

bandpasstuple of float or None, default=None

Optional analysis band (low, high) in Hz.

prominencefloat, default=0.5

Covariance-distance peak prominence multiplier.

Notes

The segmentation strategy is based on the covariance-stationarity approach used by ZapLine-plus [1].

References

[1]

Marius Klug and Niels A. Kloosterman. Zapline-plus: a zapline extension for automatic and adaptive removal of frequency-specific noise artifacts in m/eeg. Human Brain Mapping, 43(9):2743–2758, 2022. doi:10.1002/hbm.25832.

segment(data: ndarray) → list[tuple[int, int]]

Return (start_sample, end_sample) segments from channel-first data.

Parameters:

datandarray, shape (n_channels, n_times)

Input data.

Returns:

list of tuple of int

Half-open sample intervals.