Introduction
Since 2010 many companies started to incorporate biometric technology in educational platforms or services that are delivered to educational institutions \cite{Gold_2010}. One of the most common types of biometric signals that are used in research education and applications are ocular trackers. A good example of it is Tobii Technology, a high-technology company focused on developing and selling products related to eye tracking and eye control activities \cite{Gibaldi2017}. Some of their products have also been used for educational applications and research around the world \cite{Pande_2014}, in fact, his eye-tracking products have been used extensively in eye-tracking research for education. Nevertheless, most of their products require a high investment, and therefore, they are not accessible to everyone. On the other hand, Electrooculography (EOG) has been mainly used for eye tracking and eye recognition movements interfaces focused only on rehabilitation purposes or oriented for people with restricted mobility, examples of these are found in electric wheelchairs, Human-Machine Interfaces, and sleep monitoring applications \cite{Noor_2016,Usakli_2010}. Moreover, it is important to remark that electrooculography has demonstrated to be a feasible way to monitor cognitive skills that are related to the learning process such as writing, reading, and watching by monitoring only saccadic eye movements and fixation \cite{Roy_2019,Bulling_2011}. Nonetheless, the development of interfaces or devices that use EOG as a manner to study the learning process of the student and improve it, is scarce. Another drawback of the available EOG devices and interfaces is the use of specialized software and hardware to generate a high-performance recognition system that can be embedded into interfaces or applications with lower costs \cite{King2017}. This drawback is shared with the other eye-tracking devices available in the market which require a high investment on part of the professor, student, or educational institutions.
Taking the above into account, the main goal of this work is to present the design and development of an eye-tracking device for educational research purposes. The eye tracking device is made of open-source hardware and software to make it available for both students and professors and maintain it accessible for them. Therefore, this work is organized as follows: Section 2 presents a literature review of the eye-tracking studies focused on educational research and assessment. Section 3 presents the development process of the proposed eye-tracking device based on electrooculography as well as the materials and techniques used to generate the device. Section 4 shows an analysis of the learning outcomes that could be studied according to Bloom's Taxonomy with the proposed system. Section 5, presents the relationship between emotion recognition, learning, and the role or potential of electrooculography. Section 6 makes a brief comparison of the advantages and disadvantages of the proposed system in comparison to the ones used in the literature. Section 7 presents an analysis of the results. Finally, Section 8 presents the conclusions of this proposal and its further work.
Eye-tracking and education
Eye-tracking is a method that enables the gathering of information and data to make an empirical analysis of human cognition, behavior, and perception \cite{Shayan_2017}. These characteristics have enabled that over the last couple of decades, the eye-tracking devices or methods used for education applications or research related to the learning process have started to grow. Some of the fields of education in which this technology has been applied are in text reading for language learning, mathematics understanding, and science knowledge learning \cite{Zhang_2018}. Eye-tracking research focused on education has also been concerned in improving the design of computer-based learning, in the visual areas such as medicine or chess, and more recently to promote visual proficiency by eye movement modeling \cite{h2017}.
For instance, the use of eye-tracking technology has been applied to analyzed student reading behavior while reading electronic books. In the study of Kao \cite{Kao_2019}, eye-tracking technology was used to understand the reading process of students while visualizing online materials. On the other hand, image processing devices have also been used to analyze the studying behavior of mathematics content at elementary school one example of this is the work of Sun in which a software interface that employs eye-tracking technology was developed to improve the teaching framework \cite{Sun_2017}. Eye-tracking devices have started to pick up certain attention in the involvement of people that suffer from some type of disability in educational processes an example of the above is presented in the work of Moreva \cite{Moreva_2019} in which an eye-tracking platform was used to study the attention process of students while visualizing a web page-oriented for people with disabilities, in this same work it is also mentioned the growth or presence of students with some type of restricted health condition. The study of \cite{Cuesta_Cambra_2017}, tried to analyze how information is processed and learned and what is the role of visual attention in both activities. To do the above, Cuesta used eye-tracking to capture the visual behavior of the student and electroencephalography (EEG) to keep track of information processing. Moreover, Molina \cite{Molina_2018} studied the process of how electronic books are read and the importance of its design to keep the engagement of the student. To do the above, the eye movements of sixty-five students were tracked, the results indicated the importance of having a good integration of written and pictorial information in e-books. In another work of Molina \cite{manuel2017}, it is analyzed the importance of the design of multimedia materials in the process of teaching and learning. Instead of using traditional feedback from the students such as surveys and questionaries, the work of Molina employed eye-tracking since it is a more objective method to extract information related to the attention of multimedia learning materials.
According to Hajra medical educational field can also be enhanced with eye-tracking technology \cite{Ashraf_2017}. In his work, a literature review was made in which seventeen studies on clinical assessment and six focused on eye-tracking used as an assessment tool were studied, demonstrating the usefulness of the methodology in medical practices. Beach \cite{Beach_2018} shows a study in which it is determined that eye-tracking methodologies can give an insight into teachers’ engagement about learning material, reading behaviors, or sensemaking strategies. In this same study, it is remarked that these methodologies can show educational stakeholders' opportunity areas related to learning outcomes and environments. Besides, \cite{Porta_2012} provides a review of how eye-tracking technology has been employed directly or indirectly with e-learning platforms and it is described the development process of an e-learning platform to collect eye data with the intention of know the emotional state of the student. Finally, in the study of Lin \cite{Lin_2016}, 38 computer science students were monitor through eye-tracking devices while debugging C language program codes, to understand the cognitive process of the student while reviewing software. It is important to notice that most of the presented studies are mainly focused on using eye-tracking devices based on computer vision techniques, but it is not mentioned any other type of solution.
In the case of electrooculography (EOG) and its potential used in the educational field, there have been some works that have addressed the subject. For example, the study of Shipulina\cite{a2019} reports the potential that electrooculography could have educational neuroscience research in the field of mathematics. According to Shipulina EOG could provide feedback into the cognitive functions of the students by analyzing the eye-related behavior (ERB). In fact, there are studies that demonstrate that blink rates could decrease significantly during demanding cognitive tasks\cite{Fogarty1989}. Besides, Bonfiglio\cite{Bonfiglio2009} explains that blink and saccades can be used to punctuate the shifting focus of the attention of an individual from one object to another (exogenous shifts) or the reflections of one through to another (endogenous shifts). A more recent study developed by Muller \cite{M_ller_2016} intended to use the EOG technique to evaluate the usefulness of this technique in the measurement of sentence comprehension duration, the technique was compared with the traditional eye-tracking systems with the authors claiming that both techniques achieve similar results in their study. Moreover, Banerjee \cite{Banerjee_2014} tried to recognize cognitive activities form the analysis of eye movements measured through the EOG technique, eight cognitive activities were studied, activities such as reading, writing, web searching, copying, video watching, word search, relaxation activities, and game-playing were analyzed. To recognized these activities machine learning techniques were employed. Furthermore, Marandi \cite{Zargari_Marandi_2014} presented a study for the quantitative assessment of the decision-making process through the implementation of the gaze tracking by using electrooculography measurement technique, in this study subjects were asked to select between two different objects of the same class within a specific period of time. The results of this work remark that there is a statistical significance related to the gaze direction and decision-making process of the subjects.
From this brief literature review, it is possible to notice the great areas of opportunity that eye-tracking technology can have in educational applications and research related to teaching strategies. Nonetheless, most of the proposals are enabled through commercial image processing devices to collect data from the eye. In addition, there are is a lack of research or proposals that have tried to develop its own eye-tracking technologies or devices, that can serve the specific need of researchers or the experiment that is tried to be carried out by the investigators. The above led to the design of an alternative Eye-tracking technology based on electrooculography that has shown some potential used in educational research, especially in the learning cognitive process analysis. Nevertheless, the potential of eye-tracking devices based on Electrooculography has not been extendedly explored in the current state of the art, therefore it is an opportunity area that could be explored more deeply. In the next section, the process of development of the eye-tracking device is presented.
Methods
Electrooculography pre-processing circuit
Electrooculography is an electrophysiological technique in which the biopotential generated between the cornea and the retina is measure using silver/silver chloride electrodes \cite{Aungsakun_2011}. Compared with electroencephalography (EEG) techniques or Brain-Computer Interfaces (CBI), it is less invasive and requires a less quantity of electrodes. An illustration of the placement of the electrodes needed to measure the EOG signal around the eyeball can be seen in Figure \ref{809901}. With this configuration, saccadic eye movement and blinks can be easily detected through EOG and provide a measurement tool for eye-related behavior, nonetheless, there are some important problems that need to be addressed in order to perform a good eye-tracking system based on EOG. In general, this biopotential is susceptible to a great quantity of noise and variations making it rarely deterministic even for the same person in a different environment. To counterweight this problem, it is important to implement adequate processing of the signal both in hardware and software to assure a good performance of the device. The first step consists in the acquisition of the signal, then a preprocessing process is needed to remove noise and artifacts from the signal. These artifacts or noise could come from the muscle activity around the face (electromyography), the power electric line, or electromagnetic interference. The third step is related to the feature engineering of the EOG signal, this process is crucial since depending on the selected features the classification process can be done more precisely or could worsen the classification process. After the classification algorithm was selected the last step is the implementation of the whole system in an embedded processor that could process the information fo the sampled EOG signal in real-time.