Gerardo Gabriel Gentiletti

Single Trial P300 detection based on the Empirical Mode Decomposition

We present a new method for single trial detection of P300 evoked responses. The features used to classify are the coefficients of a least-squares fit of a single EEG epoch to the intrinsical mode functions of an empirical mode decomposition of the averaged event response from a P300 training set. Support vector machines with a linear kernel are used to classify the epochs and receiver operating characteristic analysis is used to evaluate our methods performance

Genetic feature selection to optimally detect P300 in brain computer interfaces

A Brain Computer Interface is a system that provides an artificial communication between the human brain and the external world. The paradigm based on event related evoked potentials is used in this work. Our main goal was to efficiently solve a binary classification problem: presence or absence of P300 in the registers. Genetic Algorithms and Support Vector Machines were used in a wrapper configuration for feature selection and classification. The original input patterns were provided by two channels (Oz and Fz) of resampled EEG registers and wavelet coefficients. To evaluate the performance of the system, accuracy, sensibility and specificity were calculated. The wrapped wavelet patterns show a better performance than the temporal ones. The results were similar for patterns from channel Oz and Fz, together or separated.

Interfaces Cerebro Computadora: Definición, Tipos y Estado Actual

The Brain Computer Interfaces (BCI) are an al- ternative of communication for people with severe motor dis- abilities. A BCI is a system that does not depend on the brains normal output pathways of peripheral nerves and muscles. These systems extract information either from EEG activity recorded from the scalp (non invasive) or the activity of indi- vidual cortical neurons recorded from implanted electrodes (invasive). In this work a synopsis of the state of the art at world-wide and national level is presented, describing the classes of BCI as well as the used paradigms for their imple- mentation.

BCI-FES system for neuro-rehabilitation of stroke patients

Nowadays, strokes are a growing cause of mortality and many people remain with motor sequelae and troubles in the daily activities. To treat this sequelae, alternative rehabilitation techniques are needed. In this article a Brain Computer Interface (BCI) system to control a Functional Electrical Stimulation (FES) system is presented. It can be used as a novel tool in easy setup clinical routines, to improve the rehabilitation process by mean of detecting patient´s motor intention, performing it by FES and finally receiving appropriate feedback The BCI-FES system presented here, consists of three blocks: the first one decodes the patient´s intention and it is composed by the patient, the acquisition hardware and the processing software (Emotiv EPOC®). The second block, based on Arduinos technology, transforms the information into a valid command signal. The last one excites the patient´s neuromuscular system by means of a FES device. In order to evaluate the cerebral activity sensed by the device, topographic maps were obtained. The BCI-FES system was able to detect the patient´s motor intention and control the FES device. At the time of this publication, the system its being employing in a rehabilitation program with patients post stroke.

Detection of Steady-State Visual Evoked Potentials based on the Multisignal Classification Algorithm

In this work we evaluated a method for detection of steady-state visual evoked potentials in one-second EEG recordings, based on the multisignal classification (MUSIC) algorithm and support vector machine classification. Three experiments were carried out to test the performance of the method and its applicability for BCI related tasks. The first experiment showed the advantages of using pseudo-spectral features derived from MUSIC over DFT-based detection, using synthetic data within a range of SNR values. A second experiment tested classification of pseudo-spectral features in a dual checkerboard stimuli condition. Finally, a third experiment with ten subjects included an additional no-stimulus condition to be detected. Results showed a faster and more accurate performance for the two- and three-class problems than previously reported DFT-based approaches.

Efecto de la Cantidad y Dimensión de los Patrones en una Interfaz Cerebro Computadora Basada en Discriminante Lineal de Fisher

The brain-computer interfaces (BCI) translate brain activity into commands for a computer. To improve the performance of BCI, it is necessary to improve the feature extraction techniques that are used to decode the intentions of the users and get a clear understanding of the basic conditions for training the classifier. In this paper we study the behavior of a linear discriminant analysis by varying the number of patterns required for training, and the number of elements used to form patterns. From the results we can conclude that for this application BCI obtain optimal performance when used about eight training patterns for each feature used. In addition, a subsample of 8 Hz value of the temporal signals of the rows of EEG showed the best overall performance as a feature extraction technique.

Filtrado mediante SVD de la onda M del electromiogramade músculos estimulados eléctricamente

The goal of this preliminary study was to investigate the feasibility of using singular value decomposition to eliminate the M-wave from the surface electromyogram (EMG) of an electrically stimulated paretic muscle in order to extract the volitional response. An SVD-based algorithm combining the subspaces method and a subsequent filtering is presented. It was evaluated with EMG signals registered from surface electrically stimulated muscles with simulated paresis and its performance was compared with a conventional fixed comb filter. A power reduction index was calculated. The filtering strategy proposed showed a good performance in static conditions where there were no traces of the M-wave. In dynamic conditions, the SVD-based algorithm was robust but with some remaining M-wave traces. It would be as a consequence of modifications in the data matrix and, therefore, in the subspaces generator columns and the singular values. In general, the fixed filter was very sensitive to input signal disturbances. In all of these conditions there was a greater power reduction for the SVD-based filter than for the fixed filter. The following step would be to evaluate the algorithm with subjects who have muscle paresis and to test it in non-controlled environments.

Detección de Potenciales Evocados en Época Única

A Single-Trial detection for Evoked Potentials is presented, it was tested with Event Related Potentials (oddball) which are used in the Donchin’s Speller. The scheme is based in inner product of registered potentials over signal subspaces, estimated by principal components analysis. Different combinations of projection coefficients were used as characteristic vectors for a gaussian Support Vector Machine classifier. To assess the scheme performance, the ROC area and accuracy were evaluated. Procedures were tested on 9 healthy subjects. Results suggest that the scheme could be used on line.