In this section, we explore and summarize the research areas that have embedded EEG-based approaches in recent years. In the past decades, an investigation of the functionality of the human brain has considerably diversified from the health domain and physiological procedures to engineering practices.
4.1. Study of Cognition
Basic research in the field of cognitive psychology persists, considering the existing gaps in knowledge, but the field continues to work in the direction of unlocking the functional mechanisms of the brain. The concept of electric interaction in the brain attracted the attention of both neurologists and cognitive psychologists, who wanted to comparatively decode normal brain functionality versus neuropathological conditions [
42,
43]. Data on wave characteristics and their effectiveness to discriminate between regional or behavioral features, for instance, gamma waves (30–70 Hz) and the involvement in conscious perception, have gradually accumulated over the years [
2]. Furthermore, brain–machine interface is on the leap, placing out to create a system capable of imitating human as closely as possible. If so, applied psychology for translational use has taken an investigative niche. Within the context of this review, many research groups have used the consumer-grade EEG gadgets in their studies, and this section of our work seeks to highlight the cutting-edge psychology-related research on emotion and cognitive functions that utilized consumer-grade EEG products, supplied by Emotiv, NeuroSky, InteraXon, and OpenBCI.
The study of cognition includes the following research areas:
- -
Emotions as emotion recognition and emotion classification;
- -
Attention and mediation;
- -
Mental workload as fatigue and stress;
- -
Memory capability.
The remainder of this section provides a broad summary of consumer-grade EEG studies on cognitive functions, together with sustained attention, working memory, and decision making.
Table 3 summarizes the prior studies reviewed in this subsection and related to the study of cognition.
Emotional processing is a popular topic for scientists, from describing how emotion is recognized, stimulated in individuals to attempting to construct emotion recognition software. A typical example is provided by [
44], which investigates the ability of the OpenBCI product in emotion recognition. This study was one of the first ones to represent evidence for the comparable performance of OpenBCI to high-end EEG derived data for emotion recognition algorithms. The data collected and analyzed with Open BCI and Empatica4 (E4) wristband led to the best prediction accuracy of 70% when using the K-means method, which indicates that the OpenBCI application could be extended to future research.
Other researchers described the usage of classification algorithms for affective tagging of audio-visual influence [
45,
46], albeit drawing EEG data using the Emotiv EPOC system. Katsigiannis and Ramzan in [
47] used the support vector machine (SVM) method for the classification of devices based on EEG features, ECG features, or a combination of both. Liu et al. in [
46] presented a real-time EEG-based emotion detection system that works on information transferred live-fed from the Emotiv EPOC headset using a 3-tier SVM-variant classification scheme. The stream through wire attaches to an EGGLAB toolbox in the Matlab for quick storage, which simplifies the signal processing, feature extraction, and machine learning implementation. The reported accuracy varies because of different tiers, ranging from approximately 65% to 92%.
As well as emotion classification, a bit different line of research explores anxiety and stress quantification [
45,
49]. One more time, both lines use EEG data and machine learning techniques advantages. Zheng et al. [
45] measured EEG and photoplethysmogram (PPG) responses in 20 volunteers using the MindWave Mobile headset and PPG-fitted glasses, respectively. To determine if EEG and PPG functions offered useful indicators for anxiety evaluation, the researchers used algorithms based on the principles of k-nearest neighbors (kNN) and SVM with radial basis function. On average, the classification accuracy was only 63%, clearly leaving some steps for improvement. Betti et al. [
49] tested an SVM algorithm on 15 separated features from physiological measurements in the Maastrict acute stress test, such as the heart rate, electrodermal, and EEG signals (MindWave Mobile), and shown an 86% accuracy in releasing stress from a relaxed state.
One of the latest studies used the system for more traditional research into regional brain activation [
48]. The participants of the experiment were involved in virtual reality (VR) simulation designed to set off sadness (e.g., poignant music, upsetting statements, or images), during which the concomitant brain activities were picked up with the Emotiv EPOC gear. The results showed that the Emotiv EPOC headset could be used as an investigative tool able to yield equivalently pertinent data to those which were received from complex neuroimaging systems.
Another field of cognition study is cognitive functions that include mental processes, e.g., attention, memory, and perception. One of such studies was using the Emotiv system and employed a combination of visual attention task and EEG brainwaves to decode attentional state to faces and scenes [
62]. After the data from the participants was collected, a linear SVM classification was addressed on the recorded signals and was demonstrated that, on average, the prediction accuracy of behavioral performance was 77%, which seems on par with the previous fMRI study. Furthermore, attention was applied in combination with mental exercise by Hashemi and colleagues from InteraXon in a large-scale investigation that compounded EEG recordings from 6029 subjects of all ages using the Muse headband [
50]. Significantly, during the investigation were noted, discernible age-related changes in brainwave characteristics, which depended on sex and frequency band. Another study, conducted in 2017, called to monitor the workers’ attention and vigilance with a wireless Emotiv product while they have been making an on-site object relocation [
59].
Besides attention, working memory is another vital element of cognitive functions. After employing the Emotiv EPOC wireless headset, a group of researchers observed focused attention and working memory of the participants while they had been completing a battery of cognitive assessment tests related to 23 fundamental cognitive skills and 5 compound skills [
60]. The extracted features from the collected EEG signals were fed to classifiers and reported that the constructed models were able to distinguish three levels (i.e., low, intermediate, and high) of focused attention and working memory with 84% and 81% accuracy, respectively. In a similar way, some other research group used a well-known dual N-back task, which was designed to tap into a person’s working memory capacity [
52]. The results showed that different memory workload levels corresponded to the reaction time of the participants and accuracy and manifested as disparate EEG patterns.
Buszard et al. in [
51] used Pearson’s automated working memory method to investigate the connection between working memory capacity, EEG coherence, and performance under pressure. The results of the experiment confirmed the correlation between verbal-analytical process, brain activity and motor performance. In total, individuals with larger verbal working memory ability exhibited greater verbal involvement (explicit motor learning) while motor performance, whereas larger visuospatial capacity, was linked to reduced verbal involvement (implicit motor learning).
Other studies of cognitive functions include research of relevance judgement (Emotiv) [
54], decision making (Emotiv) [
61], mental workload (NeuroSky) [
55], mental fatigue and stress (NeuroSky, Muse, and Emotiv) [
56,
57,
63], subconscious face recognition (Emotiv) [
53], and esthetic preference (Muse) [
58].
Ramsy et al., in [
61] investigated a person’s willingness to pay and brain activity. Recorded by Emotiv EEG signals showed that gamma and beta band asymmetry in the prefrontal areas of the brain corresponded to the willingness to pay. Frontal EEG was additionally put to harnessed for mental workload evaluation during four cognitive and motor tasks, which were calculation, finger tapping, mental rotation, and lexical decision [
55]. A concomitant increase displayed theta activity while the task was becoming more difficult, and an SVM-based model was 65% to 75% during mental workload associated with distinctive activities.
Morales et al. investigated in [
56] brain activity metrics as EEG power spectral density and saccadic peak velocity during fatigue after riding. The records were done using a ThinkGear ASIC module (TGAM) headset (NeuroSky) together with saccadic velocity sampled with infrared oculography (JAZZ-novo). To score fatigue Stanford sleepiness scale and an adapted version of the Borg rating of perceived exertion was used. The results showed that the TGAM headset could be used as a detection system for mental state modifications that occur during day-to-day task execution, for example, driving. Another recent research [
63] used Muse for detecting and predicting drowsiness by data from multiple sensors, i.e., accelerometer and gyroscope. Using a combination of the EEG spectral analysis, blinking detection from EOG, head movement from gyroscope and accelerometer, and SVM as a classifier, the researchers carried out 92% accuracy for predicting whether the participants were alert or drowsy.
The effectiveness of stress detection using EEG data was investigated in [
57,
64]. Muhlbacher-Karrer et al. in [
57] used Emotiv product to collect metrics related to a driver’s state as electrodermal activity, electrocardiogram, and capacitive hand detection sensor. The data were used to extract an average of 25 features for a channel and fed to cellular neural network classifiers (CNN). The results showed that EEG data were informative in the detection of stress, which allowed us to achieve an accuracy of 92%. The authors in [
64] used Muse to record EEG data from 28 participants in the condition of pre- and post-activity. The results showed that binary classification (stressed vs. non-stressed) with their novel feature selection algorithm had a higher prediction accuracy as compared to the 3-class alternative (93% and 64%, respectively).
Further research in the field of cognition study investigated the possibility to investigate the subconscious mind via facial recognition [
53] and to identify the individual’s artistic preference [
58]. The first work used the Emotiv EPOC headset for collecting signals during seeing of well-known people and SVM classifiers to analyze the data. Results showed an average accuracy of 65%. Second work used Muse headbands to measure EEG correlates of a group of volunteers while they had a tour of the art exhibition. The researchers found out that the piece of art that was favorite among participants caused brain activity out of the baseline condition compared to the reaction to other paintings.
In conclusion, the development of the usage psychology-related consumer-grade EEG app seems to suggest that in psychological research, using consumer-grade devices is similarly distributed between the investigation of emotion, attention, and mental processing. Additionally, in terms of the prevalent product brand, Emotiv is leading. In addition, the most interesting is that the biggest part of research mostly focused on the application of brainwave indicators as a proxy for interpretation by classification algorithms, more particularly supervised machine learning techniques.
4.2. Brain–Computer Interface
One of the major application fields for EEG equipment is the brain–computer interface (BCI). BCI has been undergoing consistent innovation and improvement since the 1970s [
65,
66]. This led to a number of products, such as the P300 speller, produced in 1998, which allowed people with motor disabilities to pick alphabets on a computer screen through visual perception and brain responses [
67]. The P300 wave is an event-related potential (ERP) event that is usually triggered by visual, auditory, and tactile stimulation. While relatively effective, the P300 speller had a prohibitive price tag; hence it never reached the consumer market. In recent years, several P300-BCI related studies have been undertaken. Some of them chose the OpenBCI platform, particularly due to its flexibility [
68,
69], others preferred the applications of Emotiv [
69,
70], and in [
71], the authors analyzed both systems. A common thread of the research was to provide more user-friendly practical applications.
Table 4 below provides a summary of the studies.
The steady-state visual evoked potential (SSVEP) represents a type of visually evoked brain response, first integrated into a BCI system in 2000 [
72]. The device aims to attribute and correlate a brain signal, as recorded by an EEG amplifier, to a specific visual stimulus frequency, such as a constant flickering stimulus on a screen. Later on, subsequent products focus on multiple frequencies and using canonical correlation analysis for discriminating between them [
73].
A limited number of studies focused on the usage of consumer-grade EEG for monitoring SSVEP-BCI. Authors in [
76] successfully linked visible feedback using a combination of Emotiv and a head-mounted device named HTC VIVE to familiarize disabled people with brain–machine interaction on a 3-D space. In two separate studies, Lamti et al. developed EEG and gaze data fusing framework for wheelchair navigation [
77] and, in [
88], Emotiv and a Tobii eye tracker (EyeX model) that were were used to monitor brain and eye activities.
Auditory steady-state responses (ASSR) are a class of steady-state EEG responses stimulated by using constant auditory stimulation (regular tone) [
89]. While a participant listens to a speaker-generated constant sound, that individual’s EEG can be varied at an equal tone or frequency (focusing on the temporal region in particular [
90]). ASSR-BCI was integrated into OpenBCI-based platforms for a home automation system allowing the user to control the devices completely with the auditory signals [
78]. Kaongoen et al. followed with a hybrid system, including ASSR and P300, also built on the OpenBCI platform [
83]. The research demonstrated that using both stimuli of ASSR and P300 simultaneously could lead to an improvement inside the performance classification of a selective attention task and reach 99% accuracy.
Motor imagery (MI) was also used as part of EEG as an innovative BCI avenue [
91]. The MI response was generated by activating the neural correlates of motor functions without actual motor execution. The most common tasks for MIBCI research are the imagery of the left/right upper and lower limbs in addition to their functional performance. Many previous studies used ultramodern experimental protocols with the datasets to conduct and analyze various imagery tasks (
http://www.schalklab.org/research/bci2000) (
http://bnci-horizon-2020.eu/).
The adaptation of the consumer-grade EEG device was mentioned in a study [
80] that used a version trained by large MI-EEG datasets got from the medical-grade device as the operating model for unseen data recorded from Emotiv. The model got acceptable results as a wheelchair controller. This research brings into focus the benefits of large-scale public datasets run from the medical-grade device in enhancing the performance of those from the consumer-grade device. To demonstrate the benefits of consumer-grade (OpenBCI), the study in [
81] collected a large amount of MI-EEG contained images right/left hand moving and resting tasks to investigate response accuracy; the study was further extended in [
82] to compare OpenBCI with medical-grade, with a DNN processing.
A different approach was taken in [
92,
93], which focused on the quality of the controls and inputs of BCI, more specifically connectivity (such as portable/wearable, wireless, dry electrodes, electrode montage, etc.), process (including online, single-channel, artifact removal especially eye blink, etc.) and environmental interaction (performed outdoors, the performance of daily life tasks, etc.).
The conclusion was that, apart from the actual strength of the BCI signals, the resulting performance is a combination of the inputs implementation, signal processing, and noise introduced by the surrounding context.
In addition to the above mentioned BCI-based studies, categorized as the research of ERP, SSVEP, and MI, a separate effort was directed towards the areas where consumer-grade EEG devices are used for both regular and medical usage. In 2018, [
84] proposed the use of the relaxation state towards the prediction of age and gender while using EEG-BCI; the results can support various applications, which include biometric, healthcare, entertainment, and targeted advertisements. The study used the Emotiv EPOC+ and achieved 88% accuracy for age and 96% for gender classification by using the random forest algorithm. In a subsequent study [
85], Kaushik et al. embedded a deep learning approach named BLSTMLSTM (a combination of bidirectional long short-term memory (BLSTM) and deep long short-term memory (LSTM)) on the same datasets and reached up to 94% and 98% for age and gender classification, respectively.
As indicated by some of the above-mentioned studies, BCI processing has recently embedded computational intelligence, deep learning (DL) to support the signal analysis and decision-making process. A new adaptation of the leading technologies can significantly reduce the cost and eliminate the mobility barrier for collecting large-scale EEG-based datasets. The cycle of data feeding into the DL algorithm can bring to the future development of applications to improve the quality of life.
4.3. Educational Research
Educational researchers prefer minimum weight, easy-to-use gadgets because of their low price, wear-resistance, single-channel, and dry electrodes. Such devices offer simpler solutions to monitor brain activity while the participants are busy with various learning tasks.
Three research areas are of substantial significance in the education-related EEG applications: attention and meditation, engagement time, and brain-to-brain synchrony.
As shown by the analysis of reviews presented in
Table 5, most of the undertaken research focused on the study of attention, in spite of the fact that there are many cognitive aspects to be explored for improving the educational environment. To further support the users, Emotiv now validates the performance metrics such as stress/frustration, engagement, interest/valence, excitement, focus/attention, and relaxation/meditation (
https://www.emotiv.com/knowledge-base/performance-metrics/). Most of the studies would also welcome a more statistically confident approach, as they tend to be based on small subject groups. Given the existing breadth of research and level of investigation, it can be concluded that educational research based on EEG is still in its early stage in terms of both technology and knowledge.
The study of attention is the most popular research area that could be seen from
Table 5. The methods for attention recognition and evaluation are mostly developed to measure the effectiveness of various teaching techniques in educational research and varied due to the feature extractions or denoising algorithms such as the usage of spectral features, principal component analysis (PCA), and multi-wavelet transform. Current studies are leaning towards machine learning as a classifier with a mixture of SVM, Bayesian classifiers, Markov models, k-nearest neighbors, and neural networks. The accuracy of the categorization is usually more than 75%, depending on the tasks and experimental installation.
In [
94], Liu et al. collected raw EEG signals from 24 people with a NeuroSky MindSet to evaluate the level of attentiveness. Totally, there were five features extracted, such as the energy value of every frequency, together with the ratio of alpha and beta activities and trained with SVM classifiers. As a result, the study reached up to 76.82% classification accuracy and showed that the alpha activity is associated with a relaxed state and the increasing of the beta activity means an increase in attention.
Since 2014, many educational studies are counting just on EEG devices as a validation tool. These studies have been using NeuroSky’s algorithm to monitor the subjects’ level of attention while they are doing the learning-related tasks. Attention level of students was investigated in [
95] using genetic algorithm for feature selection together with SVM classifier. Mobile polling, as a means of interactive learning, was studied in [
96]. The level of attention while using mobile polling may be lower compared to the traditional clicker, but it can increase during the activity. In an experiment with the affection of displayed text on the user’s attention, researchers Che and Lin [
97] have confirmed that distinct forms of text display, named static, dynamic, and mixed, do not show any tremendous results at the users’ attention. In [
98,
99], the effect of different genres of books on individuals from distinctive age groups (elementary school pupils and adults) were explored. Results showed that stimulating audiobooks or e-books are recommended to be suitable for elementary school boys, while conventional books are better for grade elementary school girls.
Additionally, to measure the maintained attention level, Chen and Wu in [
100] have exercised the emWave system to detect emotion, cognitive load, and learning performance in subjects. EmWave affords software to monitor heart rhythm and variability-based emotion recognition algorithm of heart rate. A separate direction of research, outlined in [
95,
99], used FaceReader, a facial expression detection software, to point the emotional states together with EEG-based attention level of the subjects. The results showed that the students, who obtained the rewards for the correct answers, paid more attention than the ones who received no rewards. Anyway, no connection has been determined between the positive stimuli and the learning performance.
Although the biggest part of the researchers is focusing just on attention point, some of them also include meditation into their studies as NeuroSky also supplies its pride meditation detector directly to the products [
96,
104,
106,
108]. However, meditation will not play the main role in comparison to attention during an active classroom and can only be included with other factors for performance evaluation.
Along with the usage of the consumer EEG products for evaluation of the teaching/learning methods, there are also researchers who made one more step further and developed different systems for maintaining the students’ attention and improving their learning performance. These systems usually display a user’s attention level and initiate a set of acoustic alarm when their attention drops beneath a certain threshold [
109,
110,
111]. They also can create an interactive agent that has the capacity to give proper feedback based on the user’s emotion and attention levels [
105].
The task of measuring the level of engagement is more complex than evaluating attention, may depend on the individual’s satisfaction, feeling of freshness or usability. This is one of the reasons why it is important for the learning process to have face-to-face classes or online lessons. In 2014, a new sensor-based method for analyzing the user’s motivation was proposed in [
112] by combining few metrics based on the time of interception applied by Affectiv Suite’s EmoEngine (old name of an algorithm provided by Emotiv). In addition, it is real time and can provide a deeper perception of the change in the motivation level. Authors in [
113] suggested a mechanism called time on task threshold computation (ToTCompute). The goal of this new mechanism was to help educational game creators monitoring the participant’s engagement during their interactive lessons. The system used performance metrics collected from Emotive to automatically count engagement level, and in case of low engagement value to triggering special events to motivate students.
The brain-to-brain synchrony has been developed to study social dynamics, which include the interaction among students and teachers in the real classroom and show a relationship between the brainwaves of multiple individuals related to each other. The researchers in [
114,
115] collected EEG signals from the students and their teachers in the regular classroom and computed synchrony. There were monitored brain-to-brain synchrony of the student-to-group, student-to-student, and student-to-teacher interactions. The results in [
114] showed that student-to-group synchrony was more related to teaching techniques (video or lecture), but between classmates, students showed the highest student-to-student synchrony. The results in [
115] showed that synchrony of the group of students who were studying with videos was higher than another one with lectures. The student-to-teacher synchrony showed their closeness. The results confirmed how cognitive outcomes, such as the student’s academic performance, can be predicted by social dynamics, like perceived closeness.
Other directions in investigating brain activities for education-related purposes are the classification of cognitive loads, predicting the customer frustration, the possibility estimation that the user will fail the test, detection of language comprehension and language transition in Duolingo application [
120,
121] detection of suspicious behavior during exam using EEG and eye-tracking.
The cognitive load during reading was analyzed in [
119] using Muse and EyeTribe eye-tracker for data collection. The experiment showed that brain activity during the reading of more difficult contents and easier contents is different. It displayed the influence of the textual contents on the language processing related to the left temporal region. In addition, the study analyzed the eye movement and its pattern, and the results showed the acceleration while the easier contents were reviewed, affecting the brain activity of the left frontal lobe.
The level of frustration and excitement may have an enormous influence on the students learning ability. The researchers in [
116], using Emotiv EPOC, tried to predict students’ emotion after received feedback from the Intelligent Tutoring Systems, a computer system able to provide feedback for cognitive assistance. The collected data were analyzed with several machine-learning algorithms: linear regression, kNN, and SVM to predict the user’s emotion level. The outcomes suggested that linear regression plays with higher quality than the other algorithms with R2 of 0.462, 0.560, 0.627, and 0.484 for predicting frustration, excitement, changes in frustration, and changes in excitement, respectively.
Gamification in education became very popular and identified as marketable for researchers and developers. The authors in [
118] analyzed and tried to predict the success or failure of the students in each of the tasks in the game using EEG signals from Emotiv EPOC (short/long-time excitement, mediation, frustration, and tedium), the diameter of the pupil from an eye tracker to evaluate the participants’ workload, valence, and arousal, and facial expression from FaceReader software. The game is programmed to detect if the player is progressing with the game and offer various examples or hints for better understanding. The research successfully carried out 66% accuracy with logistic regression, a machine learning model.
4.4. Gaming
From the application perspective, consumer-grade EEG devices also draw the attention of researchers in the entertainment and media sphere, especially in gaming. A number of studies, outlined in this section, aimed to investigate the ability of EEG devices to capture the feelings [
122] and the emotions [
123] of the gamers during a game by reading the activity of the brain. The recording of the brain activity has also been used to classify the level of expertise of the players involved in the game [
124] or to analyze the emotional process experienced by the patients with neurological and neurodevelopmental issues [
125]. The above works can be divided into several groups that study the:
- -
Effect on neurofunction (as emotions);
- -
Brain-controlled games;
- -
Neurological disorders and improvement.
There are many types of games, including mobile games, video games, and computer games; the games have been designed mostly for education and training, as well as serious games. It has been found that video games can relax players, bring relief from stress, improve skills, but also may be harmful and destructive to the gamers’ health and wellbeing. The outcomes of gaming can be divided into two groups, named effects of emotions and experience. The study [
122] investigated the effects of games on emotion using the Emotiv device together with saliva samples from the gamers to identify the level of stress created by the video games. Results showed that the fear and violent excitement games had increased a level of stress, and, therefore, both could have a destructive influence on the players’ health. From the other side, there was no stress after playing the puzzle game, and it was little after the runner game.
Different methods have been introduced regarding the brain functions during the playing games with the use of consumer-grade EEG devices [
122,
126].
A method suggested by Mondjar in [
127] provides the analysis of the recorded EEG signals, which showed correlation with specific mechanics of playing games, which affect the cognitive function of a player. The results showed that serious games could stimulate gamers to exercise the brain areas responsible for memory, attention, and concentration. It was determined that these cognitive activities were stimulated via five mechanics: accurate action, timely action, pattern learning, a logical puzzle, and mimic sequences.
A way to record emotions, analyze them, and determine the correlations to the signals was validated using the Emotiv EPOC headset [
122]. Compared to the previous studies, during the gameplay, the recorded brainwaves showed changes in brain activity. The results showed that various scenes during the gameplay caused differing types and intensities of emotion in each player, for example, the excitement from correctly hitting the target and frustration from missing the goal [
128]. For visualization of the brain activity have been used two colors: the area of the brain that performs high activity used red color, and the area with low activity used green color to present emotions during the game [
126,
129]. Furthermore, many neurogaming methods have funded the research of affective states on the task engagement and the 2-Dimensional valence/arousal plane [
130]. These studies aimed to help in the adaptation of various types of games for patients with neurological and neurodevelopmental disorders for clinical evaluation of emotion.
In several studies, the EEG products were used to categorize the expertise level of a player during the game based on brain activity [
124,
131,
132]. The first two works analyzed collected data and tried to predict the level of performance of each player (as professional or amateur) using naïve Bayes and SVM Classifier. Stein in [
132] launched research regarding the classification of players’ competencies and dynamic difficulty adjustment (DDA) with the help of the Emotiv system. The EEG signals were analyzed with the goal of discovering the excitement level experienced by the players. DDA means adjusting the game by reducing difficulty for weaker players and increasing difficulty for stronger players. Even more, an approach that uses the EEG signals as a base for adaptive game-based learning has been launched [
133]. The brain activity was continuously monitored, and the game mode was very fast adjusted when the decrease in the excitement level was detected in the position below a predefined threshold.
In comparison to the traditional game in which the players physically control the game avatar using a keyboard, mouse, joystick, foot paddle, brain-controlled games suggest an unconventional approach for gamers. Special headsets are capable of detecting the changes in brainwaves and give the gamers the possibility to concentrate only on the game. The idea of attention-based mind-controlled games is gaining the attention of not only game producers but also research companies [
134,
135,
136]. In 2018, Queiroz et al. [
136] launched a BCI-based game where the player has control via a wheeled robot through the Emotiv INSIGHT headset to reach a target. The headset was tuned up to software, permitting the conversion of the commands as hot or cold through the computer interface controlling the movements of the robot-like spinning around or changing direction. In the same way, Vasiljevic in [
134] and [
135] launched an attention-based BCI game named Mental War, represented as a tug-of-war game. The value of attention captured by the NeuroSky MindWave headset and read during some period of time have been calculated for both players. The higher attention value specifies the strength that use the avatar to pull the rope and win towards the opponent.
Neurological and neurodevelopmental disorders can seriously influence the patients’ everyday life, in particular on mental stability. Patients who have these types of disorders facing all types of problems related to health, academic, social relation, and occupation during their lives. Neurofeedback technology is one of the behavioral, non-pharmacological remedies, which has started to become popular because of its promising effectiveness in disorder treatment and the improvement of the patient’s health. Many studies [
125,
137,
138,
139,
140,
141,
142,
143,
144,
145,
146] presented the concept of neurofeedback training and showed that it gave a treatment option for patients, a decreasing in the fundamental symptoms such as in the cases of cerebral palsy [
147], stroke [
148], and ADHD [
149,
150,
151,
152]. Generally, the treatment consists of the patient’s request to wear a BCI headset and perform tasks specifies by a trainer. The headset records the brain signals and transforms them into visual signals for real-time feedback. For any successfully finished trial, for example, performing a suggested task correctly, there is a reward that is given for motivation to inspire the patient.
Often for treatment in conjunction with EEG headsets, the so-called “serious games” are used, a type of video games called in this way due to their special characteristics include such tasks beyond just leisure purpose, to detect any abnormal features and improve problems with attention [
129,
142]. Diverse mind-controlled games, “RehabNet” [
148], “Harvest Challenge” [
138], “Shooting” [
139], “Magic Carpet” [
143], “FOCUS” [
142], “MindLight” [
143,
144], and “FarmerKeeper” [
146], are instances of video games and serious games, which serve as neurofeedback training for patients that have cognitive issues, wherein the characters are controlled by signals changed in the brain and their patterns. All these games promote a similar goal, for example, to enhance the cognitive function of the patients with mental issues and serve as a kind of training therapy for the patients recovering from neuro-related disorders, such as ADHD, cerebral palsy, dementia, paralysis, stroke, etc. By showing the patients their brainwave patterns, it is possible to increase awareness of the specific changes in their physique that generally are not consciously controllable.
In this section, numerous factors of EEG applications have been discussed in the gaming domain: detection of emotion, gameplay media, and patient rehabilitation. and summarized in
Table 6. For now, the growing necessity of BCI for various purposes in gaming has provoked huge development of the technology. However, for future studies, it is very important to evaluate the effectiveness of the gaming system with the purpose of amplifying the data processing and hardware development for the ultimate improvement in players’ experience.