Onder Aydemir

A polynomial fitting and k-NN based approach for improving classification of motor imagery BCI data

Speed and accuracy in classification of electroencephalographic (EEG) signals are key issues in brain computer interface (BCI) technology. In this paper, we propose a fast and accurate classification method for cursor movement imagery EEG data. A two-dimensional feature vector is obtained from coefficients of the second order polynomial applied to signals of only one channel. Then, the features are classified by using the k-nearest neighbor (k-NN) algorithm. We obtained significant improvement for the speed and accuracy of the classification for data set Ia, which is a typical representative of one kind of BCI competition 2003 data. Compared with the Multiple Layer Perceptron (MLP) and the Support Vector Machine (SVM) algorithms, the k-NN algorithm not only provides better classification accuracy but also needs less training and testing times.

Decision tree structure based classification of EEG signals recorded during two dimensional cursor movement imagery

BACKGROUND Input signals of an EEG based brain computer interface (BCI) system are naturally non-stationary, have poor signal to noise ratio, depend on physical or mental tasks and are contaminated with various artifacts such as external electromagnetic waves, electromyogram and electrooculogram. All these disadvantages have motivated researchers to substantially improve speed and accuracy of all components of the communication system between brain and a BCI output device. NEW METHOD In this study, a fast and accurate decision tree structure based classification method was proposed for classifying EEG data to up/down/right/left computer cursor movement imagery EEG data. The data sets were acquired from three healthy human subjects in age group of between 24 and 29 years old in two sessions on different days. RESULTS The proposed decision tree structure based method was successfully applied to the present data sets and achieved 55.92%, 57.90% and 82.24% classification accuracy rate on the test data of three subjects. COMPARISON WITH EXISTING METHOD(S) The results indicated that the proposed method provided 12.25% improvement over the best results of the most closely related studies although the EEG signals were collected on two different sessions with about 1 week interval. CONCLUSIONS The proposed method required only a training set of the subject and automatically generated specific DTS for each new subject by determining the most appropriate feature set and classifier for each node. Additionally, with further developments of feature extraction and/or classification algorithms, any existing node can be easily replaced with new one without breaking the whole DTS. This attribute makes the proposed method flexible.

Performance evaluation of five classification algorithms in low-dimensional feature vectors extracted from EEG signals

There are lots of classification and feature extraction algorithms in the field of brain computer interface. It is significant to use optimal classification algorithm and fewer features to implement a fast and accurate brain computer interface system. In this paper, we evaluate the performances of five classical classifiers in different aspects including classification accuracy, sensitivity, specificity, Kappa and computational time in low-dimensional feature vectors extracted from EEG signals. The experiments show that naive Bayes is the most appropriate classifier for low dimensional feature vectors compared to k-nearest neighbor, support vector machine, linear discriminant analysis and decision tree classifiers.

Classification of EEG signals recorded during right/left hand movement imagery using Fast Walsh Hadamard Transform based features

Brain computer interface (BCI) allows people to communicate with machines without the use of muscle systems. Although there are various kind of techniques to understand intend of the BCI user, electroencephalography (EEG) is the most popular, practical and widely implemented one. The performance of the EEG based BCI highly depends on extracting effective features. However, there is no a general feature extraction method which provides satisfied performance for all various kind of BCI applications. Therefore, it is very valuable to develop a new feature extraction method. In this paper, we proposed a novel Fast Walsh Hadamard Transform based feature extraction method for classification of EEG signals recorded during right/left hand movement imagery. It does not only provide well-discriminative attributes but also the computational time of extracting the features from a single EEG trial is fast. The proposed method was successfully applied to Data Set III of BCI competition 2003, and achieved a classification accuracy of 88.87% on the test data. The obtained satisfactory results proved that this method can be a successful alternative to the existing feature extraction methods.

Classification of various facial movement artifacts in EEG signals

In this paper, a novel approach to classify various facial movement artifacts in EEG signals is presented. EEG signals were obtained in EEG Laboratory from three healthy human subjects in age groups between 28 and 30 years old and on different days. Extracted feature vectors based on root mean square, polynomial fitting and Hjorth descriptors were classified by k-nearest neighbor algorithm. The proposed method was successfully applied to the data sets and achieved an average classification rate of 94% on the test data.

Olfactory Recognition Based on EEG Gamma-Band Activity

There are various kinds of brain monitoring techniques, including local field potential, near-infrared spectroscopy, magnetic resonance imaging (MRI), positron emission tomography, functional MRI, electroencephalography (EEG), and magnetoencephalography. Among those techniques, EEG is the most widely used one due to its portability, low setup cost, and noninvasiveness. Apart from other advantages, EEG signals also help to evaluate the ability of the smelling organ. In such studies, EEG signals, which are recorded during smelling, are analyzed to determine the subject lacks any smelling ability or to measure the response of the brain. The main idea of this study is to show the emotional difference in EEG signals during perception of valerian, lotus flower, cheese, and rosewater odors by the EEG gamma wave. The proposed method was applied to the EEG signals, which were taken from five healthy subjects in the conditions of eyes open and eyes closed at the Swiss Federal Institute of Technology. In order to represent the signals, we extracted features from the gamma band of the EEG trials by continuous wavelet transform with the selection of Morlet as a wavelet function. Then the -nearest neighbor algorithm was implemented as the classifier for recognizing the EEG trials as valerian, lotus flower, cheese, and rosewater. We achieved an average classification accuracy rate of 87.50% with the 4.3 standard deviation value for the subjects in eyes-open condition and an average classification accuracy rate of 94.12% with the 2.9 standard deviation value for the subjects in eyes-closed condition. The results prove that the proposed continuous wavelet transform–based feature extraction method has great potential to classify the EEG signals recorded during smelling of the present odors. It has been also established that gamma-band activity of the brain is highly associated with olfaction.

A new method for selection optimum k value in k-NN classification algorithm

In this paper a new algorithm to calculate optimum value of k for k-nearest neighborhood (k-NN) is proposed. Selection of k value is very important in k-NN classification algorithm. Our algorithm applied to sub-sampling and K-fold cross validation methods, separately. We applied our algorithm in different distribution of data set with different variances and means. We compared our algorithm with other classical k selection algorithms. The results show that the proposed algorithm achieved better performance than the classical algorithms.

Classification of EEG signals recorded during right/left hand movement imagery using Fourier Transform based features

Implementation of a fast and accurate brain computer interface system depends on using a very small number of training signals, better classification and feature extraction algorithms, fewer channels and features which are improved user-specific. In this paper, we propose a fast and accurate algorithm for classifying of right/left hand movement imagery electroencephalogram data. The algorithm is presented in three basic steps. In the first step, an unit variance normalization is implemented to electroencephalogram data as preprocessing. In the second step, features are extracted from signals by using Fourier Transform algorithm. In the last step, the features are classified by using the Support Vector Machines, the k-Nearest Neighbor and Linear Discriminant Analysis. The proposed algorithm was successfully applied to the BCI competition 2003 data set which is named as Data Set III, and achieved a classification accuracy of 91.4 % on test set. The performance of the proposed algorithm was compared in terms of accuracy and speed with other studies used the same data set.

Classification of 2-dimensional cursor movement imagery EEG signals

Electroencephalogram (EEG), which is widely used for brain computer interface (BCI) systems for input signal, is easily interrupted by physical or mental tasks, and contaminated with various artifacts including power line noise, electromyogram and electrocardiogram. Therefore, such kind of artifacts cause to decrease the accuracy rate and motivate the researchers substantially develop the performance of all components of the communication system between the subject and a BCI output device. So, it is vital to use the most suitable classification algorithm and fewer feature set to implement a fast and accurate BCI system. Addition to this, it is worthwhile mentioning that the classifiers should have the ability for recognizing signals which are collected in different sessions to make brain computer interfaces practical in use. In this study, we proposed fast and accurate classification method for classifying EEG data of up/down/right/left computer cursor movement imagery. EEG signals were collected from three healthy male adults and on two different offline sessions with about one week of delay. The average test classification accuracy calculated as 53.07%.

Motor imagery ECoG signals classification using wavelet transform features

The input signals of brain computer interfaces may be either electroencephalogram (EEG) recorded from scalp or electrocorticogram (ECoG) recorded with subdural electrodes. It is very important that the classifiers have the ability for discriminating signals which are recorded in different sessions to make brain computer interfaces practical in use. This paper proposes an algorithm for classifying motor imagery ECoG signals, recorded in different sessions. Extracted feature vectors obtained with wavelet transform were classified by using k nearest neighbor method. The proposed algorithm was successfully applied to Data Set I of BCI competition 2005, and achieved a classification accuracy of 95 % on test set.