Sparse spectrum based swarm decomposition for robust nonstationary signal analysis with application to sleep apnea detection from EEG

ESSDM: An Enhanced Sparse Swarm Decomposition Method and Its Application in Multi‐class Motor Imagery–Based EEG-BCI System

“…The proposed ESSDM is a novel enhanced extension of the swarm decomposition method [24], [25] to decompose multichannel nonstationary signals into optimal OCs and improve TF analysis of nonstationary signals. The execution of the proposed ESSDM method comprises the following four major steps: Formulation of a new sparse FT spectrum, design of an iterative swarm filtering method with modification of convergence criteria, and adoption of SHO with new fitness function for optimal threshold parameter tuning that converge channel-specific optimal OCs for efficient decomposition.…”

“…Yet, the challenge of tuning multiple preset parameters across a wide range severely restricts their potential uses. Among these approaches, a newly introduced adaptive technique called swarm decomposition (SWD) [24], [25] has shown its effectiveness in improving decomposition adaptability and addressing mode mixing issues when analyzing nonstationary real signals.However, the performance of SWD relies heavily on two predefined tuning parameters: Pth (peak threshold) and StDth (iteration deviation threshold). To ensure optimal SWD decomposition, it is imperative to meticulously choose the accurate threshold settings for each signal prior to decomposition.…”

Imagined Speech-EEG Detection Using Multivariate Swarm Sparse Decomposition-Based Joint Time-Frequency Analysis for Intuitive BCI

“…The proposed MSSDM is used to design a multivariate SwF filter [16]. It is a novel multivariate extension of the sparse-swarm In contrast to mode alignment problem in the univariate SWD filtering approach [14], [15], the objective of MSSDM is to deliver kpredefined CAOCs across multichannel signals that exhibit common center frequencies and aligned compact bandwidth with common multivariate modulated oscillation.…”

“…In particular, JTFA methods adopt joint instantaneous frequency and bandwidth to model the joint oscillatory structure of multichannel signals and give enhanced joint time-frequency localization. Numerous multivariate approaches have been developed for JTFA across various real-time applications [10]- [13], Recently, an adaptive and localized TFR-based method has been implemented with the help of modified swarm decomposition (SWD) [14] and sparse Fourier transform spectrum (SFT) to enhance time-frequency localization for EEG signals [15]. Although some of these existing JTFR methods have reported significant results in time-frequency (TF) analysis, but they also introduce scope for improvement due to cross-term interference and computational cost for the multi-channel multivariate EEG signals.…”

“…In our proposed work, we have designed an end-to-end framework using the newly proposed multivariate swarm-sparse decomposition method (MSSDM) [16] to extract an ensemble of channelaligned oscillatory components (CAOCs) from multichannel IMS-EEG signals for imagined speech enhancement [15]. The proposed MSSDM explores the joint oscillatory structure of multichannel EEG signals and develops the joint TF plane for multivariate IMS-EEG signals by computing joint instantaneous frequency (JIF) and joint instantaneous amplitude (JIA) functions from signal adaptive scales.…”