brain-computer interface

Brain-computer interface systems based on classification of the motor imageries (MI) using multi-channel EEG signal play a major role in the control of artificial limbs and machines by people with disabilities. One of the main problems in classifying these signals to recognize different MI tasks is the large number of channels. The large number of channels causes a lot of cost and hassle during the measurement process, increasing computational load of the preprocessing, feature extraction, and classification, difficulty of interpretation of results, and over-fitting of the classifier due to the limited number and noisy training samples. Since not all measured channels for classifying a particular MI task have useful information, it would be beneficial to select the optimal channels for classifying desired MI tasks. Channel selection methods are categorized into wrapper, filtered, hybrid, and embedded categories. In this paper, a filtering method is used due to less computational cost and the independence of the classifier. The used criterion is very important in filtering methods. Criteria based on first- and second-order data moments are less efficient for non-Gaussian classes. The proposed method uses mutual information between candidate channels and class label as a comprehensive criterion and sequential forward selection search strategy. One of the problems in using this criterion is the accurate estimation of mutual information in the high dimensional spaces. The kpn entropy estimator is used to accurately estimate the mutual information in high dimensional space with limited number of training samples. The power of 2 Hz non overlapping sub-bands in the 8-30 Hz band and in 250 milliseconds non overlapping intervals in half to two and a half seconds after the onset of MI are extracted as features for each channel. The extracted features are reduced to 10 for each channel by combining the unsupervised L1-PCA and supervised NWFE dimensionality reduction methods. The reported results show the ability of the proposed method to select effective channels for classifying left and right hand and feet MI tasks. The overall accuracy of the SVM classifier on test samples for two subjects labeled aa and al from the BCI III competition dataset is 94.87% and 96.51%, respectively, while the number of channels is reduced from 118 to 7 channels.

  The brain-computer interface system provides a communication path between the brain and the computer and has recently received increasing attention. One of the most common paradigms of brain-computer interface systems is motor imagery. The braincomputer interface system based on motor imagery using electroencephalogram signals uses a large number of channels when receiving signals. Channels not related to the intended task cause unwanted interference and increase the noise level. In this paper, we present two optimal channel selection methods to improve common spatial pattern (CSP)-related features for classification of motor imagery tasks. Since the brain activities of motor imagery are located in a specific area of the brain, how to choose the right channels is important to improve the performance of the brain-computer interface. In this paper, analysis of variance (ANOVA) feature selection and sequential forward feature selection (SFFS) combined with common spatial pattern (CSP) are used to select optimal electrode channels. The results show that the accuracy of KNN, SVM and LDA classifications when using ANOVA+CSP method is 74, 72 and 71% respectively, when using SFFS+CSP method is 74, 73 and 68% respectively, when using CSP alone, 65, 62 and 60%, respectively, when not using the methods of selecting optimal channels and common spatial pattern, it is 58, 64 and 57%, respectively; Therefore, the combination of optimal channel selection methods and common spatial pattern has increased the accuracy of the classifiers.

Recognition of silent speech only by decoding brain signals is one of the latest research in the field of artificial intelligence. The 3rd Iranian national brain-computer interface competition, which was held by the National Brain Mapping Center of Iran in 2020, was dedicated to the classification of imagined speech for the three words of rock-paper-scissors game. In this contest, the authors introduced an approach based on wavelet packets decomposition and common spatial pattern and could win the second place. We evaluated some of the most famous classifiers including support vector machine, k-nearest neighbor, random forest, logistic regression, XGBoost and dense deep learning model separately for each subject and simultaneously for all subjects. The best average accuracy was 51.7%. Then we developed the model using spectrogram and convolutional neural network and achieved average accuracy of 76.5%. The accuracy was much better than the accuracy reported by other researchers on this dataset. Also, the performance of our model is superior to recent research in this field on different datasets.

In this study, a Brain-Computer Interface (BCI) in Silent-Talk application was implemented. The goal was an electroencephalograph (EEG) classifier for three different classes including two imagined words (Man and Red) and the silence. During the experiment, subjects were requested to silently repeat one of the two words or do nothing in a pre-selected random order. EEG signals were recorded by a 14 channel EMOTIV wireless headset. Two combinations of features and classifiers were used: Discrete Wavelet Transform (DWT) features with Support Vector Machine (SVM) classifier and Principle Component Analysis (PCA) features with a Minimum-Distance classifier. Both combinations were capable of discriminating between the three classes much better than the chance level (33.3%), none of them was reliable and accurate enough for a real application though. The first method (DWT+SVM) showed better results. In this case, feature set was D2, D3, D4 and A4 coefficients of 4-level DWT decomposition of the EEG signals, roughly corresponding to major frequency bands (Delta, Theta, Alpha and Beta) of these signals. Three binary SVM machines were used. Each machine was trained to classify between two of the three classes, namely Man/Red, Man/Silence or Red/Silence. Majority Selection Rule was used to determine final class. Once two of these classifiers presented the true class, a win (correct classification) was counted, otherwise a loss (false classification) was considered. Finally, Monte-Carlo Cross Validation showed an overall performance of about 56.8% correct classification which is comparable with the results reported for similar experiments.

The canonical correlation analysis (CCA) is one of the most widely used frequency recognition methods in steady-state visual evoked potential (SSVEP)-based brain computer interface systems. Although the CCA is often associated with good results, but if stimulation frequencies have harmonic relation, this issue will challenge this method. In this paper, the modified CCA method has been proposed that can solve this problem by adding a post-processing step in the standard CCA. For this purpose, visual stimulus ranged from 6-16 Hz with an interval of 0.5 have been generated using Matlab and the psychophysics toolbox. The SSVEP signal was recorded from ten subjects via one electrode placed at Oz. According to the proposed method, after applying CCA and determining the frequency corresponding to the maximum correlation, the difference between the correlation associated to this frequency and the correlation of the corresponding harmonic frequency is calculated. Then, the frequency is recognized by comparing the obtained value with the threshold. The threshold is determined based on the data of each subject during the offline analysis. For eight-second time window, the average recognition accuracy of the standard CCA with choosing two harmonics in constructing the reference signal (N=2) was 74%, while the corresponding value of the proposed method was 81%. Correspondingly, the accuracy was increased from 78% to 83% for four-second time window. For wide frequency range, the proposed method has been able to improve the frequency recognition accuracy compared with the standard CCA, by reducing harmonic recognition error.