20 [Finding suitable common feature sets for use in multiclass subject independent brain-21 computer interface (BCI) classifiers is problematic due to characteristically large 22 inter-subject variation of electroencephalographic signatures. We propose a wrapper 23 search method using a one versus the rest discrete output classifier. Obtaining and 24 evaluating the quality of feature sets requires the development of appropriate 25 classifier metrics. A one versus the rest classifier must be evaluated by a scalar 26 performance metric that provides feedback for the feature search algorithm. However, 27 the one versus the rest discrete classifier is prone to settling into degenerate states for 28 difficult discrimination problems. The chance of occurrence of degeneracy increases 29 with the number of classes, number of subjects and imbalance between the number of 30 samples in the majority and minority classes. This paper proposes a scalar Quality 31 (Q)-factor to compensate for classifier degeneracy and to improve the convergence of 32 the wrapper search. The Q-factor, calculated from the ratio of sensitivity to specificity 33 of the confusion matrix, is applied as a penalty to the accuracy (1-error rate). This 34 method is successfully applied to a multiclass subject independent BCI using 10 35 untrained subjects performing 4 motor tasks in conjunction with the Sequential 36 Floating Forward Selection feature search algorithm and Support Vector Machine 37 classifiers.] 3 38 1. Introduction 39 Brain Computer Interface (BCI) is an ensemble of technologies that seek to establish 40 a pathway of communication capable of translating neurologically derived signals, 41 such as imagined human movements, into computer interpretable commands [1-3]. 42 These commands can be used for the purpose of basic user interface, or for 43 controlling external devices such as a robotic arm or prosthetic [1,4-8]. BCI 44 implementation is a pattern recognition problem where signals derived from different 45 brain states are examined in order to select a set of optimally descriptive features 46 which can be most closely attributed to the associated state, typically using supervised 47 learning. 48 Many successful approaches to BCI have been presented in the literature [2,9,10]. 49 While many approaches focus on subject specific models due to high inter-subject 50 variances present in larger populations, a wide array of studies have shown the 51 viability of subject independent (SI) models [2,7,10-13]. Reducing complexity is an 52 important part of emerging consumer grade EEG devices. Viability of an off the shelf 53 subject independent solution relies on methods capable of being deployed for mobile 54 devices and embedded platforms. While the computational capabilities of these 55 devices are rapidly increasing, power requirements play a key role in feasibility of 56 high computational complexity models. As a result, optimized models based on 57 traditional machine learning techniques gain an advantage for BCI applications. 58 Additional...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.