Aida Khorshidtalab

Robust classification of motor imagery EEG signals using statistical time–domain features

The tradeoff between computational complexity and speed, in addition to growing demands for real-time BMI (brain-machine interface) systems, expose the necessity of applying methods with least possible complexity. Willison amplitude (WAMP) and slope sign change (SSC) are two promising time-domain features only if the right threshold value is defined for them. To overcome the drawback of going through trial and error for the determination of a suitable threshold value, modified WAMP and modified SSC are proposed in this paper. Besides, a comprehensive assessment of statistical time-domain features in which their effectiveness is evaluated with a support vector machine (SVM) is presented. To ensure the accuracy of the results obtained by the SVM, the performance of each feature is reassessed with supervised fuzzy C-means. The general assessment shows that every subject had at least one of his performances near or greater than 80%. The obtained results prove that for BMI applications, in which a few errors can be tolerated, these combinations of feature-classifier are suitable. Moreover, features that could perform satisfactorily were selected for feature combination. Combinations of the selected features are evaluated with the SVM, and they could significantly improve the results, in some cases, up to full accuracy.

A Transform-Based Feature Extraction Approach for Motor Imagery Tasks Classification

In this paper, we present a new motor imagery classification method in the context of electroencephalography (EEG)-based brain-computer interface (BCI). This method uses a signal-dependent orthogonal transform, referred to as linear prediction singular value decomposition (LP-SVD), for feature extraction. The transform defines the mapping as the left singular vectors of the LP coefficient filter impulse response matrix. Using a logistic tree-based model classifier; the extracted features are classified into one of four motor imagery movements. The proposed approach was first benchmarked against two related state-of-the-art feature extraction approaches, namely, discrete cosine transform (DCT) and adaptive autoregressive (AAR)-based methods. By achieving an accuracy of 67.35%, the LP-SVD approach outperformed the other approaches by large margins (25% compared with DCT and 6 % compared with AAR-based methods). To further improve the discriminatory capability of the extracted features and reduce the computational complexity, we enlarged the extracted feature subset by incorporating two extra features, namely, Q- and the Hotellings

ECG Parametric Modeling Based on Signal Dependent Orthogonal Transform

T^{2}

Evaluation of time-domain features for motor imagery movements using FCM and SVM

statistics of the transformed EEG and introduced a new EEG channel selection method. The performance of the EEG classification based on the expanded feature set and channel selection method was compared with that of a number of the state-of-the-art classification methods previously reported with the BCI IIIa competition data set. Our method came second with an average accuracy of 81.38%.

Motor Imagery Task Classification Using a Signal-Dependent Orthogonal Transform Based Feature Extraction

In this letter, we propose a parametric modeling technique for the electrocardiogram (ECG) signal based on signal dependent orthogonal transform. The technique involves the mapping of the ECG heartbeats into the singular values (SV) domain using the left singular vectors matrix of the impulse response matrix of the LPC filter. The resulting spectral coefficients vector would be concentrated, leading to an approximation to a sum of exponentially damped sinusoids (EDS). A two-stage procedure is then used to estimate the model parameters. The Pronys method is first employed to obtain initial estimates of the model, while the Levenberg-Marquardt (LM) method is then applied to solve the non-linear least-square optimization problem. The ECG signal is reconstructed using the EDS parameters and the linear prediction coefficients via the inverse transform. The merit of the proposed modeling technique is illustrated on the clinical data collected from the MIT-BIH database including all the arrhythmias classes that are recommended by the Association for the Advancement of Medical Instrumentation (AAMI). For all the tested ECG heartbeats, the average values of the percent root mean square difference (PRDs) between the actual and the reconstructed signals were relatively low, varying between a minimum of 3.1545% for Premature Ventricular Contractions (PVC) class and a maximum of 10.8152% for Nodal Escape (NE) class.

Evaluating the effectiveness of time-domain features for motor imagery movements using SVM

Brain-Machine Interface is a direct communication pathway between brain and an external electronic device. BMIs aim to translate brain activities into control commands. To design a system that translates brain waves and its activities to desired commands, motor imagery tasks classification is the core part. Classification accuracy not only depends on how capable the classifier is but also it is about the input data. Feature extraction is to highlight the properties of signal that make it distinct from the signal of the other mental tasks. Performance of BMIs directly depends on the effectiveness of the feature extraction and classification algorithms. If a feature provides large interclass difference for different classes, the applied classifier exhibits a better performance. In order to attain less computational complexity, five time-domain procedure, namely: Mean Absolute Value, Maximum peak value, Simple Square Integral, Willison Amplitude, and Waveform Length are used for feature extraction of EEG signals. Two classifiers are applied to assess the performance of each feature-subject. SVM with polynomial kernel is one of the applied nonlinear classifier and supervised FCM is the other one. The performance of each feature for input data are evaluated with both classifiers and classification accuracy is the considered common comparison parameter.

Motor imagery task classification using transformation based features

In this paper, we present the results of classifying electroencephalographic EEG signals into four motor imagery tasks using a new method for feature extraction. This method is based on a signal-dependent orthogonal transform, referred to as LP-SVD, defined as the left singular vectors of the LPC filter impulse response matrix. Using a logistic tree based model classifier, the extracted features are mapped into one of four motor imagery movements, namely left hand, right hand, foot, and tongue. The proposed technique-based classification performance was benchmarked against those based on two widely used linear transform for feature extraction methods, namely discrete cosine transform DCT and adaptive autoregressive AAR. By achieving an accuracy of 67.35i¾ź%, the LP-SVD based method outperformed the other two by large margins +25i¾ź% compared to DCT and +6i¾ź% compared to AAR-based methods.

Testing and analysis of the proposed data driven method on the opportunity human activity dataset

Motor imagery electroencephalogram signals are the only bio-signals that enable locked-in patients, who have lost control over every motor output, to communicate with and control their surroundings. Brain Machine Interface is collaboration between a human and machines, which translates brain waves to desired, understandable commands for a machine. Classification of motor imagery tasks for BMIs is the crucial part. Classification accuracy not only depends on how accurate and robust the classifier is; it is also about data. For well separated data, classifiers such as kernel SVM can handle classification and deliver acceptable results. If a feature provides large interclass difference for different classes, immunity to random noise and chaotic behavior of EEG signal is rationally conformed, which means the applied feature is suitable for classifying EEG signals. In this work, in order to have less computational complexity, time-domain algorithms are employed to motor imagery signals. Extracted features are: Mean Absolute Value, Maximum peak value, Simple Square Integral, Willison Amplitude, and Waveform Length. Support Vector Machine with polynomial kernel is applied for classification of four different classes of data. The obtained results show that these features have acceptable, distinct values for different these four motor imagery tasks. Maximum classification accuracy belongs to contribution of Willison amplitude as feature and SVM as classifier, with 95.1 percentages accuracy. Where, the lowest is the contribution of Waveform Length and SVM with 31.67 percentages classification accuracy.

Classification of Retinal Images Based on Statistical Moments and Principal Component Analysis

Abstract This paper proposes a feature extraction method named as LP_QR, based on the decomposition of the LPC filter impulse response matrix of the signal of interest. This feature extraction method is inspired by LP_SVD and is tested in the context of motor imagery electroencephalogram. The extracted features are classified and benchmarked against extracted features of LP_SVD method. The two applied methods are also compared regarding the required execution time, which further highlights their respective merits and demerits. This paper closely examines the contribution of EEG channels of these two information extraction algorithms too. Consequently, a detailed analysis of the role of EEG channels concerning the nature of the extracted information is presented. This study is conducted on the BCI IIIa competition database of four motor imagery movements. The obtained results indicate that the proposed method is the better choice if simplicity is demanded. The investigation into the role of EEG channels reveals that level of contribution each channel can be quite dissimilar for different feature extraction algorithms.

Modeling of Human Arm Movement: A Study on Daily Movement

This paper proposes a data-driven method for constructing materials to be used in a probabilistic knowledge base for human activity recognition. The utilized dataset, challenge subset of Opportunity, is a publicly available dataset. It consists of a set of daily activities, which has been manually labeled as modes of locomotion and gestures. We applied several methods to extract proper features from sensors on bodies of subjects, then, chosen features are fed into two different classifiers. Finally, predicted labels for modes of locomotion and hand gestures are calculated. To evaluate the method, the recognition rates are benchmarked against the results of the competitors who have participated in Opportunity challenge as well as the baseline results provided by the Opportunity group. For modes of locomotion, our results surpass all of the available results and in some cases the recognition rate of our model is very close to the highest recognition rate. For gestures, regular or noisy data, in some cases our method is still higher than baseline or challenge participants but unlike locomotion, it is not capable to beat them all.