Abstract
Superior feature extraction, channel selection and classification methods are essential for designing electroencephalography (EEG) classification frameworks. However, the performance of most frameworks is limited by their improper channel selection methods and too specifical design, leading to high computational complexity, non-convergent procedure and narrow expansibility. In this paper, to remedy these drawbacks, we propose a fast, open EEG classification framework centralized by EEG feature compression, low-dimensional representation, and convergent iterative channel ranking. First, to reduce the complexity, we use data clustering to compress the EEG features channel-wise, packing the high-dimensional EEG signal, and endowing them with numerical signatures. Second, to provide easy access to alternative superior methods, we structurally represent each EEG trial in a feature vector with its corresponding numerical signature. Thus, the recorded signals of many trials shrink to a low-dimensional structural matrix compatible with most pattern recognition methods. Third, a series of effective iterative feature selection approaches with theoretical convergence is introduced to rank the EEG channels and remove redundant ones, further accelerating the EEG classification process and ensuring its stability. Finally, a classical linear discriminant analysis (LDA) model is employed to classify a single EEG trial with selected channels. Experimental results on two real world brain-computer interface (BCI) competition datasets demonstrate the promising performance of the proposed framework over state-of-the-art methods.
Highlights
For a long time, the brain-computer interface (BCI) has been a prevalent communication system for directly bridging between a brain and a computer
We provide an EEG classification framework applicable to most common EEG processing and pattern recognition methods to provide an entrance for increasing the EEG-based BCI (EEG-BCI) performance through technology integration
We refer to our methods based on robust feature selection (RFS), robust unsupervised feature selection (RUFS), and SSLSR as FCCR1, FCCR2, and FCCR3, respectively, and compare them with some baselines
Summary
The brain-computer interface (BCI) has been a prevalent communication system for directly bridging between a brain and a computer. A typical BCI system collects brain activities associated with mental tasks, translates these neural signals into appropriate commands, and eventually sends them to a computer (Handiru and Prasad, 2016). As a non-invasive brain activity measurement method, electroencephalography (EEG) has attracted increasing interest, owing to its low risk, low cost, feasibility, and significant potential for practical applications (Yang et al, 2016). Feature extraction is fundamental to EEG classification and many methods have been presented to acquire the implicit essential information from raw EEG signals, e.g., spectral power (SP) and time-domain parameters (TDP) (Müller et al, 2004). Channel selection attempts to remove irrelevant EEG features in redundant channels to reduce the setup time and equipment cost, and improve the effectiveness and efficiency of EEG-BCI systems (Alotaiby et al, 2015). Well-known channel selection methods are often based on evolutionary algorithms and mutual information
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have