mne_denoise.ssa.LocalSingularSpectrumAnalysis#

class mne_denoise.ssa.LocalSingularSpectrumAnalysis(window_length: int | None = None, *, window_seconds: float | None = None, sfreq: float | None = None, n_clusters: int | str = 'auto', max_clusters: int = 10, max_window: int = 100, random_state: int | None = 0, verbose: bool | str | int | None = None)[source]#

Local-SSA high-amplitude artifact transformer.

The estimator applies the clustered local-subspace reconstruction of Teixeira et al. independently to each selected channel and exposes the fitted clustering diagnostics after transformation.

Parameters:

  • window_length (int | None, default=None) – Delay-vector dimension in samples. It is mutually exclusive with window_seconds. None selects an automatic value.

  • window_seconds (float | None, default=None) – Delay-vector duration in seconds. It requires a sampling frequency and is mutually exclusive with window_length.

  • sfreq (float | None, default=None) – Sampling frequency in Hz. MNE input supplies it from metadata and must agree with an explicit value.

  • n_clusters (int | "auto", default="auto") – Number of delay-vector clusters, or automatic reliable selection.

  • max_clusters (int, default=10) – Upper bound for automatic cluster-count selection.

  • max_window (int, default=100) – Maximum delay-vector dimension used by automatic window selection.

  • random_state (int | None, default=0) – Random seed passed to k-means.

  • verbose (bool | str | int | None, default=None) – MNE-style logging level.

sfreq_#

Validated sampling frequency, or None when sample-based parameters and NumPy input do not require one.

Type:

float | None

n_channels_in_#

Number of data channels seen during fitting.

Type:

int

ch_names_in_#

Fitted MNE channel names and order, or None for NumPy input.

Type:

tuple of str | None

diagnostics_#

Diagnostics from the most recent transformation. Epoched input stores one dictionary per epoch.

Type:

dict | list of dict

n_clusters_#

Effective cluster count per channel, or per epoch and channel.

Type:

ndarray

subspace_dimensions_#

Selected local subspace dimensions for every channel.

Type:

list

See also

compute_local_ssa

Functional interface for channel-first arrays.

local_ssa_clean_channel

Canonical single-channel implementation.

mne_denoise.ssa.SingularSpectrumAnalysis

Basic SSA with frequency grouping.

Notes

The estimator is transductive. fit validates the operating point and records the channel layout; every transform clusters and decomposes the records supplied to that call. Record and epoch boundaries can therefore change the delay vectors, clusters, covariance spectra, and reconstruction.

Local SSA assumes that coherent, high-energy structure is artifact. Genuine neural activity that satisfies the same local-subspace model can be removed [1].

References

[1]

Teixeira, A. R., Tome, A. M., Lang, E. W., Gruber, P., & Martins da Silva, A. (2006). Automatic removal of high-amplitude artefacts from single-channel electroencephalograms. Computer Methods and Programs in Biomedicine, 83, 125-138. https://doi.org/10.1016/j.cmpb.2006.06.003

Examples

>>> import numpy as np
>>> from mne_denoise.ssa import LocalSingularSpectrumAnalysis
>>> sfreq = 100.0
>>> time = np.arange(500) / sfreq
>>> data = np.vstack(
...     [np.sin(2 * np.pi * 0.5 * time), np.sin(2 * np.pi * 10.0 * time)]
... )
>>> model = LocalSingularSpectrumAnalysis(
...     window_length=20, n_clusters=2, random_state=0
... )
>>> cleaned = model.fit_transform(data)
>>> cleaned.shape
(2, 500)
__init__(window_length: int | None = None, *, window_seconds: float | None = None, sfreq: float | None = None, n_clusters: int | str = 'auto', max_clusters: int = 10, max_window: int = 100, random_state: int | None = 0, verbose: bool | str | int | None = None) → None[source]#

Methods

__init__([window_length, window_seconds, ...])

fit(X[, y])

Validate the operating point and record the fitted channel layout.

fit_transform(X[, y])

Fit to data, then transform it.

get_metadata_routing()

Get metadata routing of this object.

get_params([deep])

Get parameters for this estimator.

set_output(*[, transform])

Set output container.

set_params(**params)

Set the parameters of this estimator.

transform(X[, y])

Apply the transductive SSA decomposition to the supplied records.