Hybrid Wavelet and EMD/ICA Approach for Artifact Suppression in Pervasive EEG

Abstract: Electroencephalogram (EEG) signals are often corrupted with unintended artifacts which need to be removed for extracting meaningful clinical information from them. Typically a priori knowledge of the nature of the artifacts is needed for such purpose. Artifact contamination of EEG is even more prominent for pervasive EEG systems where the subjects are free to move and thereby introducing a wide variety of motion-related artifacts. This makes hard to get a priori knowledge about their characteristics rendering conventional artifact removal techniques often ineffective. In this paper, we explore the performance of two hybrid artifact removal algorithms: Wavelet packet transform followed by Independent Component Analysis (WPTICA) and Wavelet Packet Transform followed by Empirical Mode Decomposition (WPTEMD) in pervasive EEG recording scenario, assuming existence of no a priori knowledge about the artifacts and compare their performance with two existing artifact removal algorithms. Artifact cleaning performance has been measured using Root Mean Square Error (RMSE) and Artifact to Signal Ratio (ASR) - an index similar to traditional Signal to Noise Ratio (SNR), and also by observing normalized power distribution topography over the scalp. Comparison has been made first using semi-simulated signals and then with real experimentally acquired EEG data with commercially available 19-channel pervasive EEG system Enobio corrupted by eight types of artifact. Our explorations show that WPTEMD consistently gives best artifact cleaning performance not only in semi-simulated scenario but also in the case of real EEG data containing artifacts.

• The paper presents hybrid techniques WPTEMD and WPTICA that effectively remove motion-induced artifacts from pervasive EEG data.
• Using multiscale analysis, it employs Wavelet Packet Transform with either EMD or ICA to isolate and discard artifact components.
• Experimental evaluations show up to 70.4% improvement in artifact suppression metrics and enhanced signal fidelity for real-world applications.

Hybrid Wavelet and EMD/ICA Approach for Artifact Suppression in Pervasive EEG

Introduction and Motivation

Electroencephalography (EEG) signals are crucial for numerous applications, including cognitive function assessment and neuro-disease diagnosis. However, EEG data is often contaminated with artifacts, especially in pervasive and unconstrained environments where motion-induced artifacts are prevalent. Traditional artifact removal algorithms require a priori knowledge about the artifacts, which is impractical in dynamic EEG settings. This paper introduces two novel hybrid techniques designed to tackle this problem: the Wavelet Packet Transform followed by Empirical Mode Decomposition (WPTEMD) and Wavelet Packet Transform followed by Independent Component Analysis (WPTICA). These methods aim to effectively suppress motion-related artifacts in pervasive EEG without relying on predefined artifact characteristics.

Methods and Techniques

Wavelet Packet Transform (WPT)

The WPT is employed as the initial step in both hybrid techniques to decompose the EEG signals into different frequency components, enabling an effective analysis and identification of corrupted segments. This transform allows for the multiscale analysis of EEG signals, maintaining high frequency resolution and aiding in localizing artifacts.

Empirical Mode Decomposition (EMD) and Independent Component Analysis (ICA)

In the WPTEMD approach, following WPT, the EMD is applied to discard Intrinsic Mode Functions (IMFs) associated with artifacts, characterized by irregular oscillations and high entropy. Conversely, the WPTICA utilizes ICA to identify and remove independent components linked to artifacts. This method capitalizes on the statistical independence properties of EEG signal components.

Experimental Evaluation

The performance of WPTEMD and WPTICA was rigorously evaluated using both semi-simulated and real EEG datasets, gathered through a commercially available 19-channel pervasive EEG system, Enobio. The semi-simulated approach allowed a controlled comparison against benchmark methods like wICA and FASTER, establishing a performance baseline with quantified signal recovery, measured by Root Mean Square Error (RMSE) and Artifact to Signal Ratio (ASR).

Numerical Results

WPTEMD outperformed other examined methods, showing a 49.53% and 70.4% improvement in RMSE and ASR for heavily contaminated channels. In scenarios with real EEG data encompassing eight types of artifacts—ranging from blinking to various head movements—WPTEMD consistently achieved superior artifact suppression without distorting the underlying EEG spectral characteristics. This superior performance was particularly evident in low-density EEG channel setups that posed significant challenges for existing methods.

Discussion

The hybrid nature of WPTEMD and WPTICA allows for flexibility and robustness in artifact suppression, crucial for pervasive EEG applications where movement artifacts are common and unpredictable. The adaptive nature of EMD in WPTEMD, specifically the empirical extraction and suppression of IMFs, demonstrates notable advantages in maintaining EEG signal integrity across varying artifact intensities and types.

Conclusion and Future Work

This study showcases the efficacy of WPTEMD as a robust artifact suppression tool in pervasive EEG scenarios, free from reliance on prior artifact knowledge. Its application improves EEG signal fidelity and utility in dynamic and naturalistic assessment settings. Future work may focus on refining these hybrid strategies further by integrating more sophisticated machine learning-based classification techniques to enhance automatic artifact identification and suppression in ultra-low-density EEG systems.

This work has broad implications for the deployment of portable, real-time EEG monitoring systems in naturalistic environments, enabling more accurate monitoring of cognitive states and potential medical applications outside of traditional lab settings.