A High Performance Spelling System based on EEG-EOG Signals With Visual Feedback

Abstract

In this paper, we propose a highly accurate and fast spelling system that employs multi-modal electroencephalography-electrooculography (EEG-EOG) signals and visual feedback technology. Over the last 20 years, various types of speller systems have been developed in brain-computer interface and EOG/eye-tracking research; however, these conventional systems have a tradeoff between the spelling accuracy (or decoding) and typing speed. Healthy users and physically challenged participants, in particular, may become exhausted quickly; thus, there is a need for a speller system with fast typing speed while retaining a high level of spelling accuracy. In this paper, we propose the first hybrid speller system that combines EEG and EOG signals with visual feedback technology so that the user and the speller system can act cooperatively for optimal decision-making. The proposed spelling system consists of a classic row-column event-related potential (ERP) speller, an EOG command detector, and visual feedback modules. First, the online ERP speller calculates classification probabilities for all candidate characters from the EEG epochs. Second, characters are sorted by their probability, and the characters with the highest probabilities are highlighted as visual feedback within the row-column spelling layout. Finally, the user can actively select the character as the target by generating an EOG command. The proposed system shows 97.6% spelling accuracy and an information transfer rate of 39.6 (±13.2) [bits/min] across 20 participants. In our extended experiment, we redesigned the visual feedback and minimized the number of channels (four channels) in order to enhance the speller performance and increase usability. Most importantly, a new weighted strategy resulted in 100% accuracy and a 57.8 (±23.6) [bits/min] information transfer rate across six participants. This paper demonstrates that the proposed system can provide a reliable communication channel for practical speller applications and may be used to supplement existing systems.

Towards Brain-Computer Interfaces for Drone Swarm Control

“…For the acquired EEG data verification, we adopted a basic EEG classification procedure which is generally used conventional BCI studies [1]- [4]. The data were preprocessed by using a band-pass filter with a zero-phase 2nd Butterworth filter between Hz.…”

“…Brain-computer interface (BCI) analyzes brain signals to understand intention and status of human that can be used for controlling various machines. Since brain signals contain significant information about status of human, many BCI studies have attempted to understand brain signals [1]- [4]. In contrast, invasive methods such as electrocroticogram (ECoG) [5] place the electrodes on the brain directly to acquire high-quality brain signals.…”

“…EEG-based BCI has several paradigms for signal acquisition such as motor imagery (MI) [6]- [8], event-related potential (ERP) [9], [10], and movement-related cortical potential (MRCP) [2], [11]. As applications of EEGbased BCI, a robotic arm [6], [12], [13], a speller [4], [14]- [16], a wheelchair [17], and a drone [18]- [21] were commonly used for communication between human and machines. Recently, a BCI-based drone swarm is one of the most interesting topics ( Fig.…”

Classification of Upper Limb Movements Using Convolutional Neural Network with 3D Inception Block

“…Most non-invasive BMI-based robotic arm systems have been designed using different types of electroencephalography (EEG) modality [9]- [12] such as a steady-state visually evoked potential (SSVEP) [13]- [15], an event-related potential (ERP) [16]- [18], error-related potential (ErrP) [19], movement imagery (MI) [20], [21] and movement execution (ME) [22], [23]. In this paper, we focused on ME paradigm in which can design a robotic arm system.…”

Prediction of Individual Propofol Requirements based on Preoperative EEG Signals

“…EEG is one of the methods to measure brain activity [11][12][13][14][15]. Unlike other methods such as functional magnetic resonance imaging (fMRI) and functional near-infrared spectroscopy (fNIRS), which obtain the hemodynamic signal, EEG measures differences of electrical potential on the scalp [16].…”