Figure 2 - uploaded by Sara Åsly
Content may be subject to copyright.
Source publication
This thesis has investigated the possibility to classify EEG signals produced by visual exposure to red, green and blue (RGB) colors, thought to provide rapid control and decreased learning times brain-computer interface (BCI) applications. An in-house experiment with 17 participants was designed and conducted. Analytic and empirical signal analysi...
Similar publications
The research aims to reduce or eliminate the training stage when a new subject uses a BCI (Brain-Computer Interface) based on imagined speech. This article presents a series of experiments that seek to verify if your training information already available for a group of subjects can be used for a new subject. The method uses self-organized maps to...
![Fig. 1 Headset Emotiv Insight. [5] There are five EEG sensors to sense...](publication/336009502/figure/fig1/AS:808751632224259@1569832823909/Headset-Emotiv-Insight-5-There-are-five-EEG-sensors-to-sense-brain-activity-and-two_Q320.jpg)
![Fig. 2 Locations of seven EEG sensors. [5]](publication/336009502/figure/fig2/AS:808751632232456@1569832823967/Locations-of-seven-EEG-sensors-5_Q320.jpg)
Brain-Computer Interface (BCI) is a very attractive area and common trend worldwide. The electroencephalographic (EEG) technology is being used by BCI systems that process brain signals through computer algorithms. EEG-based BCI has become an important tool for real-time analysis of brain activity. This article examines the usability and quality of...
Citations
... In [72], SSVEP was used to flash three colors, and the results showed that it was possible to classify the EEG signals to a color. Two master thesis's [73,74] showed promising results for classifying three colors when the subjects were exposed to continuous colored light. Classification between color exposure and restingstate has also been proved feasible by the study [75], where the best results were created with a SVM classifier. ...
This thesis investigates the feasibility of a simple communication system for persons with Locked-in syndrome (LIS) by using a combination of the brain’s color perception and the eye movement of the user. A person diagnosed with LIS is conscious and awake but trapped in his/her own body, unable to move and communicate. The communication system proposed here consists of a brain-computer interface (BCI) that uses recorded electroencephalography (EEG) signals generated after a dedicated visual stimulation protocol.
The BCI design needs a classification model, and this thesis explores different state-of-the-art pro- cessing and classification methods for the EEG signal. The classification task is split into two prob- lems. The first problem consists of differentiating between a task state where the subject looks at a presented color and a resting state. The second problem consists of differentiating between the vari- ous task states, a subject looking at one of four different colors. An in-house experiment was designed and conducted to create a dataset that fits the designed BCIs specifications. The dataset includes recorded data from 22 healthy subjects, where everyone was exposed to two different protocols. The first protocol alternated between exposing the participants to one of four colors and a resting state. The second protocol displayed the color with a superimposed background icon indicative of a user- oriented need.
The results from the experiments showed that the proposed methods predicted similarly well on in- put data from both protocols. A random forest (RF) classifier proved to predict best on average when trained and tested on data from just one subject. The results calculated from the 22 individual RF models reached the average accuracies of 74.3 % and 61.4 % for differentiating between a task and resting state and between the four task states, respectively. RF reached these results by decomposing the input signal with variational mode decomposition (VMD), where the fractals, energies, and sta- tistical features extracted from the modes were used.
Finally, a general model that could predict task-related information from new subjects was tested. The best performing model was a state-of-the-art convolutional neural network (CNN). The model was pre-trained on data from an optimized selection of subject data from a new dataset by the non- dominated sorting genetic algorithm II (NSGA-II). Then, the model performed a short calibration of its weights on 60 % of the data from the new subject the model was going to predict. The average accuracy for differentiating between a task and resting state and between the four task states was 69.8 % and 73.6 %, respectively. This demonstrates that a general model, only needing to calibrate on a few new samples from the user, can be used to create a BCI communication system.
... They observed that the difference in frequency response is a good classification Equal contribution signature. In [10] the intrinsic mode functions (IMFs) for Empirical mode decomposition (EMD) were used to identify features in the brain signals that describe the colour activity. The IMFs were used as input to classifiers such as Random Forest (RF) and Naive Bayes. ...
... Of these, LDAs gave the highest average accuracy of 67.07%. The accuracy obtained when classifying RGB-colours in this paper are higher than the accuracy obtained in [10] and [5] with an average accuracy of 46% and 70.2% respectively. However, the equipment for recording this dataset used gel based electrodes and impedance was controlled, contrary to [10] and [5], where dry electrodes were used. ...
... The accuracy obtained when classifying RGB-colours in this paper are higher than the accuracy obtained in [10] and [5] with an average accuracy of 46% and 70.2% respectively. However, the equipment for recording this dataset used gel based electrodes and impedance was controlled, contrary to [10] and [5], where dry electrodes were used. ...
Identifying unique descriptors for primary colours in EEG signals will open the way to Brain-Computer Interface (BCI) systems that can control devices by exposure to primary colours. This study is aimed to identify such unique descriptors in visual evoked potentials (VEPs) elicited in response to the exposure to primary colours (RGB: red, green, and blue) from 31 subjects. For that, we first created a classification method with integrated transfer learning that can be suitable for an online setting. The method classified between the three RGB classes for each subject, and the obtained average accuracy over 23 subjects was 74.48%. 14 out of 23 subjects were above the average level and the maximum accuracy was 93.42%. When cross-session transfer learning was evaluated, 86% of the subjects tested showed an average variation of 5.0% in the accuracy comparing with the source set.
... Some of the proposed work has already been carried out on di erent EEG signal classi cation tasks. For example, a similar process was used in a Master's degree theses [310][311][312] and the same process for feature extraction and classi cation of the response to RGB color exposure [313][314][315]. The process for channel selection using NSGA-II was also used for source localization, reducing the number of EEG channels from 231 to less than 10, while obtaining similar localization errors [316]. ...
The aim of this thesis is to move one step forward towards the concept of electroencephalographic (EEG) systems that can achieve the same objectives as high-density EEG with a minimum required number of channels. This requires EEG signal analysis, computational intelligence, and optimization techniques that can systematically identify the minimum number of channels that fulfills the objectives currently achieved with high-density EEG systems. Achieving this goal will pave the way towards the hardware-software realization of user-centric, easy-to-use, readily affordable EEG systems for universal applications. Enabling portability while ensuring performance of comparable or higher quality than that of high-density EEG will expand the accessibility of EEG to non-traditional users and personal applications moving EEG out of the lab. The application horizon will be expanded from experimental research to clinical use, to the gaming industry, intelligence and security sectors, education and daily use by people for self-knowledge.
The methods proposed in the thesis comprise the combination of feature extraction techniques and channel selection algorithms with optimization techniques that allow extracting the most essential information from a minimum set of required EEG channels that were tested in two cases-studies: Epileptic seizure classification, and EEG-based biometric systems. The Discrete Wavelet Transform (DWT) and Empirical Mode Decomposition (EMD) were used to decompose EEG signals into different frequency bands and then four features were computed for each sub-band, the Teager and Instantaneous energies and the Higuchi and Petrosian fractal dimensions.
For the optimization stage, non-dominated sorting genetic algorithms (NSGA) were used for channel selection, using binary values to represent the channels in the chromosomes, $1$ if the channel is used in the classification and optimization process, and $0$ if not. Additional genes to represent important parameters for the classifiers were added using integer and decimal values.
For Case-study 1, NSGA-III selected one or two channels from a set of 22 for epileptic seizure classification, obtaining an accuracy of up to 0.98 and 1.00, respectively, using EMD/DWT-based features.
For Case-study 2, a task-independent, resting-state-based biometric system using Local Outlier Factor (LOF)- and DWT-based features showed a True Acceptance Rate (TAR) of up to 0.993±0.01 and a True Rejection Rate (TRR) of up to 0.941±0.002 using only three channels selected by NSGA-III from a set of 64.
The results presented herein can be considered to be a first proof-of-concept, showing that it is possible to reduce the number of required EEG channels for classification tasks and opens the way to explore these methods on other neuroparadigms. This will lead to reduced real-time computational costs for EEG signal processing, removing task-irrelevant and redundant information, as well as reducing the preparation time for use of the EEG headsets.
The results of such a reduction in the number of required EEG channels will make possible a low-power hardware design, expanding the range of EEG-based applications from clinical diagnosis and research to health-care, to non-medical applications that can improve our understanding of cognitive processes, learning and education and to the discovery of current hidden/unknown properties behind ordinary human activity and ailments.
... Recently, the use of portable headband devices have gained popularity due to their ease of use and accessibility. The authors in [2,16] have used four channel G.tec's MOBIlab four channel portable device in a problem similar to ours. However, their method yielded a lower accuracy of 58% in comparison to our proposed approach. ...
... They are indicators of statistical properties used in signal processing in the time domain introduced by Bo Hjorth in 1970 [28]. We obtained 16 Hjorth parameters for alpha and beta band for all 4 channels. We calculated two Hjorth parameters namely, the mobility parameter as in equation 2 and complexity parameter as in equation 3 on alpha and beta power bands that we obtain from CWT. ...
The perception of color is an important cognitive feature of the human brain. The variety of colors that impinge upon the human eye can trigger changes in brain activity which can be captured using electroencephalography (EEG). In this work, we have designed a multiclass classification model to detect the primary colors from the features of raw EEG signals. In contrast to previous research, our method employs spectral power features, statistical features as well as correlation features from the signal band power obtained from continuous Morlet wavelet transform instead of raw EEG, for the classification task. We have applied dimensionality reduction techniques such as Forward Feature Selection and Stacked Autoencoders to reduce the dimension of data eventually increasing the model's efficiency. Our proposed methodology using Forward Selection and Random Forest Classifier gave the best overall accuracy of 80.6\% for intra-subject classification. Our approach shows promise in developing techniques for cognitive tasks using color cues such as controlling Internet of Thing (IoT) devices by looking at primary colors for individuals with restricted motor abilities.
... Aiming to design a novel visual BCI, some works have approached, with different degree of success, the use of the EEG responses to either color exposure [2]- [6] or the imagination of colors [3], [4], [7]. These systems are independent of a flickering stimulator and they could take advantage of the fact that color-based cues are already a part of our daily life such as traffic lights, allowed/forbidden access doors, etc. ...
... However, it is not clear enough if this could be achieved in color-exposure-based BCIs. The most related work, described in [2], presented a preliminary experiment to distinguish between RGB colors and idle state using a pre-processed version of the dataset recorded in [3]. The outcomes were computed generating a single dataset with all the instances of all RGB color exposure (as a single class) and idle state from seven subjects and then training and testing several classifiers such as random forest, support vector machines (SVM), K-nearest neighbors (KNN), decision tree, and Naive Bayes. ...
... These feature vectors were then used in the classification stage, for the assessment of performance, using four classifiers after the application of 10fold cross-validation. It is important to highlight that we will use the same method as in [2] (described above), aiming to assess the pertinence of this method for the proposed questions in this paper. Specific details are provided below for each stage of the method. ...
Brain-Computer Interfaces are systems that would allow either the reintegration of handicapped people or effective control of devices by thoughts for healthy people. Several neuroparadigms (motor imagery, SSVEPs, P300, SCP among others) have been explored to control a BCI, but passive color exposure is one of the most novels. However, the analysis of the potential of this task for being used in a self-paced BCI has not been studied enough. In this paper, the first step towards a self-paced color-exposure-based BCI is presented. This is done by the analysis of the differences between the interest activities (RGB colors) and idle state. Specifically, we extract EMD (Empirical Mode Decomposition)-based features from the EEG signals and then these are classified using 4 classifiers (SVM, random forest, Naive Bayes and KNN). We then present an analysis of the performances for these two classes (RGB colors vs idle state) obtained for seven subjects, highlighting SVM's average performance (93.27% for all of them). Also, the differences between the two classes (color exposure and idle state) suggest that they do not depend on a specific subject. Last, we provide experimental evidence that there is no a specific color that could be distinguished better regardless of idle state, reaching average accuracies of 85.01%, 85.56%, and 86.91% for red, green and blue vs idle state, respectively.
Self-paced Brain-Computer Interfaces (BCIs) are desirable for allowing the BCI's user to control a BCI without a cue to indicate him/her when to send a command or message. As a first step towards a self-paced color-based BCI, we assessed if a machine learning algorithm can learn to distinguish between primary color exposure and idle state. In this paper, we record and analyze the EEG signals from 18 subjects for assessing the feasibility of distinguishing between color exposure and idle states. Specifically, we compare separately the performances obtained in the classification of two different types of idle states (one relaxation-related and another attention-related) and color exposure. We characterize the signals using two different ways based on discrete wavelet transform and Empirical Mode Decomposition (EMD), respectively. We trained and tested two different classifiers, support vector machine (SVM) and random forest. The outcomes provide experimental evidence that a machine learning algorithm can distinguish between the two classes (exposure to primary colors and idle states), regardless of the kind of idle state analyzed. The more consistent outcomes were obtained using EMD-based features with accuracies of 92.3% and 91.6% (considering a break and an attention-related task as the idle states). Also, when we discard the epochs' onset the performances were 91.8% and 94.6%, respectively.
