A Novel Deep Learning Scheme for Motor Imagery EEG Decoding Based on Spatial Representation Fusion

Jun Yang,Yunfa Fu,Jin Wang,Zhengmin Ma

doi:10.1109/access.2020.3035347

Abstract

Motor imagery electroencephalography (MI-EEG), which is an important subfield of active brain-computer interface (BCI) systems, can be applied to help disabled people to consciously and directly control prosthesis or external devices, aiding them in certain daily activities. However, the low signal-to-noise ratio and spatial resolution make MI-EEG decoding a challenging task. Recently, some deep neural approaches have shown good improvements over state-of-the-art BCI methods. In this study, an end-to-end scheme that includes a multi-layer convolution neural network is constructed for an accurate spatial representation of multi-channel grouped MI-EEG signals, which is employed to extract the useful information present in a multi-channel MI signal. Then the invariant spatial representations are captured from across-subjects training for enhancing the generalization capability through a stacked sparse autoencoder framework, which is inspired by representative deep learning models. Furthermore, a quantitative experimental analysis is conducted on our private dataset and on a public BCI competition dataset. The results show the effectiveness and significance of the proposed methodology.

Highlights

Brain–computer interface systems (BCIs) [1,2,3] try to map human intention from brain activities, providing a new pathway between the human brain and the external environment
A convolution neural network (CNN) based on different granular-grouped channels and a stacked sparse autoencoder (SSAE)-combined CNN method were applied to MI data for improving the discriminative and generalization capability of the Motor imagery electroencephalography (MI-EEG) decoding model
The datasets collected from our laboratory and from BCI Competition IV dataset 1 were used

Summary

Introduction

Brain–computer interface systems (BCIs) [1,2,3] try to map human intention from brain activities, providing a new pathway between the human brain and the external environment. For EEGbased BCI studies, motor imagery electroencephalography (MI-EEG), which is the only active BCI paradigm without the requirement of external stimuli, is a popular and key research topic in BCI applications. The key step in MI-BCI tasks is to decode the MI-EEG signals efficiently. The EEG-based BCIs is regarded as a pipeline framework, which includes three main parts: 1) Signal pre-processing involving data augmentation, noise and artefact removal, and electrode channel selection; 2) Feature extraction and representation of the appropriate properties or subcomponents of the constructed signal; 3) Classification that involves outputting the discrimination result by decoding the brain intention. Conventional techniques employ machine learning approaches with handcrafted features for decoding EEG signal [12,13].

Objectives

Methods

Findings

Conclusion