Objective: Artifact subspace reconstruction (ASR) is an automatic, online-capable, component-based method that can effectively remove transient or large-amplitude artifacts contaminating electroencephalographic (EEG) data. However, the effectiveness of ASR and the optimal choice of its parameter have not been systematically evaluated and reported, especially on actual EEG data. Methods: This study systematically evaluates ASR on twenty EEG recordings taken during simulated driving experiments. Independent component analysis (ICA) and an independent component classifier are applied to separate artifacts from brain signals to quantitatively assess the effectiveness of the ASR. Results: ASR removes more eye and muscle components than brain components. Moreover, even though some eye and muscle components retain after ASR cleaning, the power of their activities are reduced. Study results also showed that ASR cleaning improved the quality of a subsequent ICA decomposition. Conclusions: Empirical results show that the optimal ASR parameter is between 20 and 30, balancing between removing non-brain signals and retaining brain activities. Significance: With an appropriate choice of parameter, ASR can be a powerful and automatic artifact removal approach for offline data analysis or online real-time EEG applications such as clinical monitoring and brain-computer interfaces.

Brain–computer interface (BCI) actively translates the brain signals into executable actions by establishing direct communication between the human brain and external devices. Recording brain activity through electroencephalography (EEG) is generally contaminated with both physiological and nonphysiological artifacts, which significantly hinders the BCI performance. Artifact subspace reconstruction (ASR) is a well-known statistical technique that automatically removes artifact components by determining the rejection threshold based on the initial reference EEG segment in multichannel EEG recordings. In real-world applications, the fixed threshold may limit the efficacy of the artifact correction, especially when the quality of the reference data is poor. This study proposes an adaptive online ASR technique by integrating the Hebbian/anti-Hebbian neural networks into the ASR algorithm, namely, principle subspace projection ASR (PSP-ASR) and principal subspace whitening ASR (PSW-ASR) that segmentwise self-organize the artifact subspace by updating the synaptic weights according to the Hebbian and anti-Hebbian learning rules. The effectiveness of the proposed algorithm is compared to the conventional ASR approaches on benchmark EEG dataset and three BCI frameworks, including steady-state visual evoked potential (SSVEP), rapid serial visual presentation (RSVP), and motor imagery (MI) by evaluating the root-mean-square error (RMSE), the signal-to-noise ratio (SNR), the Pearson correlation, and classification accuracy. The results demonstrated that the PSW-ASR algorithm effectively removed the EEG artifacts and retained the activity-specific brain signals compared to the PSP-ASR, standard ASR (Init-ASR), and moving-window ASR (MW-ASR) methods, thereby enhancing the SSVEP, RSVP, and MI BCI performances. Finally, our empirical results from the PSW-ASR algorithm suggested the choice of an aggressive cutoff range of ${c =}$ 1–10 for activity-specific BCI applications and a moderate range of ${c > }10$ for the benchmark dataset and general BCI applications.

Artifact Subspace Reconstruction (ASR) is a machine learning technique widely used to remove non-brain signals (referred to as “artifacts”) from electroencephalograms (EEGs). The ASR algorithm can, however, be constrained by the limited memory available on portable devices. To address this challenge, we propose a Hardware-Oriented Memory-Limited Online ASR (HMO-ASR) algorithm. The proposed HMO-ASR algorithm consists of (1) two-level window-based preprocessing including PCA-based and z-score-based preprocessing to clean the data in each window, (2) iterative mean, standard deviation, and covariance update using a parallel algorithm to achieve window-based processing, and (3) early eigenvector matrix determination to save the computation. With the three schemes, the HMO-ASR method can be implemented on mobile devices, application-specific integrated circuits (ASICs) or field programmable gate arrays (FPGAs) with limited memory. The study results showed that the proposed HMO-ASR algorithm can achieve comparable performance to those obtained by the offline ASR algorithm with a 98.64% reduction in memory size. An FPGA implementation is used for silicon proof of the proposed HMO-ASR algorithm.