A Generalized Model to Estimate Reaction Time Corresponding to Visual Stimulus Using Single-Trial EEG Mohammad Samin Nur Chowdhury 1* , Arindam Dutta 1 , Matthew K. Robison 2 , Chris Blais 2 , Gene Brewer 2 , and Daniel W. Bliss 1 Abstract— The estimation of the visual stimulus-based reaction time (RT) using subtle and complex information from the brain signals is still a challenge, as the behavioral response during perceptual decision making varies inordinately across trials. Several investigations have tried to formulate the estimation based on electroencephalogram (EEG) signals. However, these studies are subject-specific and limited to regression-based analysis. In this paper, for the first time to our knowledge, a generalized model is introduced to estimate RT using single-trial EEG features for a simple visual reaction task, considering both regression and classification-based approaches. With the regression-based approach, we could predict RT with a root mean square error of 111.2 ms and a correlation coefficient of 0.74. A binary and a 3-class classifier model were trained, based on the magnitude of RT, for the classification approach. Accuracy of 79% and 72% were achieved for the binary and the 3-class classification, respectively. Limiting our study to only high and low RT groups, the model classified the two groups with an accuracy of 95%. Relevant EEG channels were evaluated to localize the part of the brain significantly responsible for RT estimation, followed by the isolation of important features. Clinical relevance— Electroencephalogram (EEG) signals can be used in Brain-computer interfaces (BCIs), enabling people with neuromuscular disorders like brainstem stroke, amyotrophic lateral sclerosis, and spinal cord injury to commu- nicate with assistive devices. However, advancements regarding EEG signal analysis and interpretation are far from adequate, and this study is a step forward. I. INTRODUCTION The human brain controls a person’s actions/reactions, and response to a particular event directly depends on the type of stimulus and the brain functionalities. A lot of previous psychological experiments on human behavior have been conducted to understand human reactions in terms of brain activities [1]. However, as the brain is one of the largest and most complicated parts of the human body with uncountable functions, there is a long way to go to unfold all the mysteries. This paper concentrates on the human reaction towards visual stimulus and tries to estimate reaction time (RT) from information embedded in electroencephalogram (EEG) signals. A recent study [2] demonstrated a method to estimate RT based on Riemannian tangent space features from EEG This work was not supported by any organization 1 M. S. N. Chowdhury, A. Dutta, and D. W. Bliss are with School of Electrical, Computer & Energy Engineering, Arizona State University 2 M. K. Robison, C. Blais, and G. Brewer are with Dpt. of Psychology, Arizona State University * Corresponding author email: mchowd12@asu.edu data taken from 16 participants. Another investigation [3] tried to estimate the timing of target onset during rapid serial visual presentation based on single-trial EEG using data from 6 subjects. There are other similar studies, and all of them employ subject-specific, regression-only estimation models. This study tries to overcome the existing limitations by proposing a feature extraction-based generalized model that uses single-trial EEG data. Also, it takes into account a classification approach aside from regression. Multi-domain features were extracted from different EEG channels and fed to machine learning models for the estimation purpose. Based on the outcomes, channels (brain locations) and fea- tures that are highly related to the response were separated. II. METHODS A. Experiment Details A total of 48 participants (31 males, 17 females) of ages ranging from 18 to 24 performed a simple visual reaction task. The experimental setup can be explained as follows. In a computer screen, a plus symbol (+) appears at the center, and after variable time duration, the symbol changes into a cross (×). The task of every subject is to tap the space bar as soon as the symbol changes from plus (+) to cross (×). This action is repeated several times for each subject, and every single repetition is considered as a trial. For each subject, the duration of the experiment was about 30 minutes, and the average number of trials was more than 310. During the entire experiment, noninvasive scalp EEG signals from 30 effective channels were recorded for all the participants. The subjects had normal or corrected to normal vision with no history of neurological problems, and informed consent was obtained from each of them. The Institutional Review Board of Arizona State University approved all the procedures. B. Data Handling and Preprocessing The data had a sampling rate of 1kHz and were filtered from DC to 400 Hz to get rid of the high-frequency noise. The ocular artifacts were removed by visual inspection of the ocular components obtained from Independent component analysis (ICA). Afterward, the data were normalized by mean and variance. In the EEG data, there was a pair of starting and ending point indicators for each subject. The following events repeated several times sequentially between them depending on the number of trials: 1) beginning of the trial (+) 2) change of symbol (×) 3) response of the subject (space bar tap)