A Thresholding-Free State Detection Approach for Home Appliance Control Using P300-Based BCI

Abstract

The main objective of brain-computer interface (BCI)-based asynchronous system is to detect user's intentional control (IC) and non-control (NC) states. In IC state, users are constantly adapting to the BCI system and making selections, whereas in NC state, users ignore to focus on the BCI system. Most of the current methods for IC or NC state detection are based on thresholding. However, choosing a suitable threshold is often time-consuming. Moreover, the chosen threshold may be insufficient over a long period of time due to the variability of EEG signals. This article proposes an asynchronous P300-based row-column paradigm (RCP) system for home appliances control. The proposed system is able to detect IC and NC states of the subject. A novel 2×3 matrix consisting of home appliances (such as light, fan, mobile device, door, television, and electric heater) has been proposed as an RCP. For stimulation, technical paper & video clip were shown on a computer screen to the users. Step-wise linear discriminant analysis (SWLDA) classifier was employed in the proposed thresholding-free asynchronous-BCI system to distinguishes between user IC and NC states. In the proposed experiment, fourteen healthy and nine paralyzed persons have participated as volunteer subjects. The synchronous mode average classification accuracy of 92.44% and 89.33% was achieved for healthy and paralyzed participants, respectively. Thereafter, the average asynchronous mode accuracy of 94.28% and 92.11% with very less false activation rate (FARs) of 0.15/min and 0.162/min was achieved for healthy and paralyzed participants, respectively. As an application, the developed P300-based speller was used to control home appliances by issuing different commands. The experimental results validated the effectiveness of the proposed method and presents future direction to use the system for paralyzed patients.