Scale-Dependent Signal Identification in Low-Dimensional Subspace: Motor Imagery Task Classification.

Qingshan She,Tom Potter,Yingchun Zhang,Zhizeng Luo,Haitao Gan,Yuliang Ma

doi:10.1155/2016/7431012

Abstract

Motor imagery electroencephalography (EEG) has been successfully used in locomotor rehabilitation programs. While the noise-assisted multivariate empirical mode decomposition (NA-MEMD) algorithm has been utilized to extract task-specific frequency bands from all channels in the same scale as the intrinsic mode functions (IMFs), identifying and extracting the specific IMFs that contain significant information remain difficult. In this paper, a novel method has been developed to identify the information-bearing components in a low-dimensional subspace without prior knowledge. Our method trains a Gaussian mixture model (GMM) of the composite data, which is comprised of the IMFs from both the original signal and noise, by employing kernel spectral regression to reduce the dimension of the composite data. The informative IMFs are then discriminated using a GMM clustering algorithm, the common spatial pattern (CSP) approach is exploited to extract the task-related features from the reconstructed signals, and a support vector machine (SVM) is applied to the extracted features to recognize the classes of EEG signals during different motor imagery tasks. The effectiveness of the proposed method has been verified by both computer simulations and motor imagery EEG datasets.

Highlights

Many people throughout the world live with a variety of clinical conditions, including stroke, spinal trauma, cerebral palsy, and multiple sclerosis
The new algorithm was constructed based on the spectral regression code and the Gaussian mixture model (GMM) clustering code found in the software package
The noise-assisted MEMD (NA-MEMD) algorithm was first performed on original datasets to obtain a set of multivariate intrinsic mode functions (IMFs), with the subsequent application of unsupervised kernel spectral regression (KSR) to generate low-dimensional feature vectors by mapping the decomposed IMFs into lower-dimensional subspace

Summary

Introduction

Many people throughout the world live with a variety of clinical conditions, including stroke, spinal trauma, cerebral palsy, and multiple sclerosis These conditions frequently present with motor deficits, which greatly reduce the quality of life for those affected. Despite recording noninvasively and on the same time scale as the sensorimotor control of the brain, the high-dimensional EEG data used in MI exercises faces many challenges [4]. These signals are usually collected from multiple electrodes (or channels), which are inevitably contaminated by the noise from biological, environmental, and instrumental origins

Objectives

Methods

Results

Discussion

Conclusion