Classification of intended motor movement using surface EEG ensemble empirical mode decomposition

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Noninvasive electroencephalography (EEG) brain-computer interface (BCI) systems are crucial for assisting individuals with motor impairments.
• Investigating intended arm reaching tasks is key to developing effective prosthetic control schemes.
• Current BCI systems require enhanced accuracy for reliable prosthetic device operation.

Purpose of the Study:

• To develop and evaluate a BCI control scheme for prosthetic devices.
• To improve control over prosthetic devices for individuals with limited motor control.
• To assess classification accuracy using different signal processing and machine learning techniques.

Main Methods:

• Utilized noninvasive electroencephalography (EEG) to record brain activity during intended arm reaching tasks.
• Employed Independent Component Analysis (ICA) for artifact identification and EEGLAB for validating post parietal cortex (PPC) activation.
• Evaluated binary classification strategies including Support Vector Machine (SVM) and Fisher Linear Discrimination (FLD) on surface electrode signals near PPC.

Main Results:

• Linear Fisher Linear Discrimination (FLD) showed comparable accuracy (63.41%) to nonlinear Support Vector Machine (SVM) (63.65%).
• Introducing a normalization factor based on visual cue 'signature' significantly boosted classification accuracy.
• Accuracy reached 90.43% for raw signals and 93.55% for intrinsic mode functions (IMF) using Ensemble Empirical Mode Decomposition (EEMD).

Conclusions:

• Standard classification methods showed limited improvement for BCI control.
• Signal normalization techniques, particularly with EEMD, offer a significant pathway to enhance BCI accuracy.
• This approach holds promise for improving prosthetic device control for individuals with motor disabilities.