decoding_ssd.h File Reference

Spatio-Spectral Decomposition (SSD) for noise-aware narrowband enhancement of continuous M/EEG. More...

#include "decoding_global.h"
#include <Eigen/Core>

Include dependency graph for decoding_ssd.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
class	DECODINGLIB::DecodingSsd
	Spatio-spectral decomposition for narrowband signal enhancement on continuous M/EEG. More...

Namespaces
namespace	DECODINGLIB
	Supervised and unsupervised spatial-filter decompositions for M/EEG decoding.

Detailed Description

Spatio-Spectral Decomposition (SSD) for noise-aware narrowband enhancement of continuous M/EEG.

SPDX-License-Identifier: BSD-3-Clause Copyright (c) 2026 MNE-CPP Authors

Author

Christoph Dinh chris.nosp@m.toph.nosp@m..dinh.nosp@m.@mne.nosp@m.-cpp..nosp@m.org

Since

2.2.1

Date

May 2026

SSD extracts spatial filters that maximise variance in a narrow signal band of interest while simultaneously minimising variance in the flanking noise band. It does so by solving the generalised eigenvalue problem \(\Sigma_s w = \lambda \Sigma_n w\), where \(\Sigma_s\) is the covariance of the data band-passed to the signal band and \(\Sigma_n\) is the covariance of the data band-passed to the adjacent noise band. The result is a small set of components with the highest achievable signal-to-noise ratio for the band of interest, which is what makes SSD the standard preprocessor for narrowband oscillation studies (alpha, mu, beta peaks). The method was introduced by Nikulin, Nolte & Curio, A novel method for reliable and fast extraction of neuronal EEG/MEG oscillations on the basis of spatio-spectral decomposition, NeuroImage 55, 2011.

DecodingSsd mirrors mne.decoding.SSD and adds an explicit DecodingSsd::apply method that reconstructs the sensor-space signal from a chosen subset of components, providing a low-rank denoised version of the input that the connectivity and inverse pipelines can consume without re-filtering. The covariance of the noise band is shrunk towards a scaled identity by regParam to keep the GED well-conditioned in the typical high-channel-count / short- recording regime. Bandpassing is performed internally with a zero- phase windowed-sinc FIR (forward + reverse pass) so the same call site works regardless of the upstream filtering history.

Definition in file decoding_ssd.h.