Development of robust electrooculography (EOG)-based human-computer interface controlled by eight-directional eye movements

Abstract

Electrooculography (EOG) signal is one of the useful electro-physiological signals. The EOG signals provide information about eye movements that can be used as a control signal in human-computer interface (HCI). Usually, eight-directional movements, including up, down, right, left, upright , up-left, downright and down-left, are proposed. Development of the EOG signal classification has been shown more increasing interest in the last decade; however, the effect of noises on classification system is a major problem to degrade the usefulness of EOG-based HCI. A robust classification algorithm of the eight movements is proposed, in which this technique can conduct the effect of noises in EOG signal, particularly for involuntary movements and eye-blink artifacts. The proposed algorithm was based on the onset analysis, feature extraction, the first derivative technique and threshold classification. Eight beneficial time domain features were proposed including the peak and the valley amplitude positions, and the upper and the lower wavelengths of two EOG channels, vertical and horizontal channels. Based on the optimal threshold values and conditions, the results showed that classification accuracy reached 100% for three-subject testing. In addition, the first derivative technique was additionally implemented in order to avoid the eye-blink artifact and other eight time domain features, that is, peak amplitude and area under curve, have been investigated for use in advanced HCI interfaces, notably, eye activity and eye writing recognitions.

Leveraging Deep Neural Networks for Lucid Dream Communication via Two-Dimensional Electrooculography

“…In our Pilot Study 1, the proposed CNN-based eye gaze estimation achieved a general accuracy of 77.6%, demonstrating potential under low data quality conditions in a sleeping setting with participants lying down and eyes closed. Static threshold methods(Park, Kim, & Kim, 2005;Aungsakun, Phinyomark, Phukpattaranont, & Limsakul, 2012) exhibit the highest average accuracy of 97.5%, matching the accuracy of our proposed static threshold implemented in Pilot Study 2.While fixed threshold analysis shows higher accuracy compared to machine learning methods, it remains untested whether these methods are equally effective for detecting eye gestures composed of sequences of saccades, such as eye typing gestures and commands in Pilot Study 1. Convolutional Neural Networks (CNNs) hold potential in classifying noisy sequences of saccades and offer higher signaling speed without the need for dwell time to separate eye movements.…”

“…For instance, eye tracking interfaces enable communication through eye commands (Kim, Han, & Im, 2018;Heo, Yoon, & Park, 2017) and eye typing (Majaranta & Räihä, 2002). Detection of eye movements in these studies employs techniques including EOG channel threshold analysis (Aungsakun, Phinyomark, Phukpattaranont, & Limsakul, 2012;Tsai, Lee, Wu, Wu, & Kao, 2008), Support Vector Machines (Bulling, Ward, Gellersen, & Tröster, 2010), and Deep Neural Networks (Fang & Shinozaki, 2018), among others. A comprehensive summary is available in (Ramkumar, Kumar, Rajkumar, Ilayaraja, & Shankar, 2018).…”

Development of a Magnetoresistive-Based Wearable Eye-Tracking System for Oculomotor Assessment in Neurological and Otoneurological Research—Preliminary In Vivo Tests

“…The proposed device needs only a preliminary self-calibration of the sensors, after which it produces absolute angle estimations for the ambient field and for the magnet (some calibration can be necessary to convert magnet orientation to eye-gaze orientation); moreover, its functioning does not suffer from light changes or blinking, differently from IROG (Infrared OculoGraphy), which requires a patient-based recalibration and is very sensitive to external light changes, so that environment light changes can produce some biases during the data acquisition procedure [33]. IROG consists of an IR light source that illuminates the eye and an array of photodetectors that collect the reflected light [34,35], thus it cannot produce any signal when the eyes are closed and suffers from artefacts driven by eye-blinking. On the other hand, the magnetic tracker is more invasive compared to other optical devices (IROG, IR cameras) but it is not influenced by blinking.…”

Nine States HCI using Electrooculogram and Neural Networks

“…A robust classification algorithm based on onset analysis, first derivative technique and threshold analysis were proposed by the authors. Based on the optimal threshold values and conditions, the result showed that classification accuracy reached 100% for three subjects during testing [20]. Most of the research on EOG based HCI focuses only on limited states varying from two to six.…”