Mostrar el registro sencillo del documento
Time series analysis using information measures supporting supervised learning tasks
dc.rights.license | Atribución-SinDerivadas 4.0 Internacional |
dc.rights.license | Atribución-SinDerivadas 4.0 Internacional |
dc.contributor.advisor | Castellanos Domínguez, César Germán |
dc.contributor.author | Luna-Naranjo, David Felipe |
dc.date.accessioned | 2020-03-02T19:33:12Z |
dc.date.available | 2020-03-02T19:33:12Z |
dc.date.issued | 2019 |
dc.identifier.uri | https://repositorio.unal.edu.co/handle/unal/75781 |
dc.description.abstract | The brain-computer interfaces provide an alternative control of the devices through the activity of the human brain. The selection of channels as a stage for the development of BCI systems allows to evaluate the important information and improve the performance of the systems. However, the current methodologies for channel selection are not always exact, they also have a high computational cost because in the acquisition of the data redundant data are presented. For this reason, a methodology that is robust and has an optimal performance in the selection of channels is necessary. This in order to reduce costs in the acquisition, reduce the computational cost and improve overall performance. This document proposes three methods to improve the classification performance in BCI tasks.\\ The first is a trial-wise channel filtering by selecting the subset of independent components with the largest entropy. This method holds two free parameters: The order for the Renyi entropy weighs the component quantization according to its probability, and the percentage of retained entropy that rules the number of independent components to reconstruct the spatially filtered EEG channels. Both free parameters are tuned using a subject-dependent grid search for the best classification accuracy. The results show that using ICA as a spatial filtering allows the feature extraction stage to build more discriminating spaces, reducing the influence of non-informative components. As an advantage, the resulting spatial filtering maintains the physiological interpretation of the EEG channels. The second method is a relevance analysis based on the maximum mean discrepancy as the distance function between a pair of single-channel trials, termed rMMD. The proposed rMMD starts with a trial embedding that highlights temporal dynamics, and ends with a channel ranking according to a designed relevance function. The function relies on the within and between class distances to quantify the discrimination capability of each channel.In comparison with no channel selection and a heuristic approach, our proposed relevance analysis statistically improves the classification of MI tasks with a reduced set of channels. We evaluate the rMMD on a bi-class motor-imagery (MI) dataset holding 64 channels and more than 40 subjects. The third method proposes a measure of relevance based on the projection of the data in a high-dimensional space from the kernel trick method in which correntropy is used as a measure of similarity. This allows the method to be robust to the variations temporary and variability present in the different tests. From this measure, a selection of patterns is made that significantly improves the classification performance in bi-class motor imagery task (MI) in a public database composed of 22 channels and nine subjects. |
dc.description.abstract | Las interfaces cerebro-computadora proporcionan un control alternativo de los dispositivos a través\ de la actividad del cerebro humano. La selección de canales como etapa para el desarrollo de sistemas BCI permite evaluar la información importante y mejorar el desempeño de los sistemas. Sin embargo las metodologías actuales para realizar la selección de canales no son siempre exactas, ademas tienen un alto coste computacional debido a que en la adquisición de los datos se presentan datos redundantes. Por esta razón es necesaria una metodología que sea robusta y tenga un optimo desempeño en la selección de canales. Esto con la finalidad de disminuir costos en la adquisición, reducir el coste computacional y mejorar el desempeño general. Este documento propone tres métodos para mejorar el rendimiento de clasificación en tareas de BCI. El primero es un filtrado de canales seleccionando el subconjunto de componentes independientes con la entropía más grande. Este método tiene dos parámetros libres: el orden para la entropía de Renyi que pesa la cuantificación del componente de acuerdo con su probabilidad, y el porcentaje de entropía retenida que gobierna el número de componentes independientes para reconstruir los canales EEG espacialmente filtrados. Ambos parámetros libres se ajustan usando una búsqueda de cuadrícula dependiente del sujeto para obtener la mejor precisión de clasificación. Los resultados muestran que el uso de ICA como un filtro espacial permite que la etapa de extracción de características cree espacios más selectivos, reduciendo la influencia de los componentes no informativos. Como ventaja, el filtrado espacial resultante mantiene la interpretación fisiológica de los canales EEG. El segundo método es un análisis de relevancia basado en la discrepancia media máxima como la función de distancia entre un par de ensayos de un solo canal, denominado rMMD. La rMMD propuesta comienza con un mapeo de las pruebas que destaca la dinámica temporal y finaliza con una clasificación de canales según una función de relevancia diseñada. La función depende de las distancias internas y entre clases para cuantificar la capacidad de discriminación de cada canal. En comparación con ninguna selección de canales y un enfoque heurístico, nuestro análisis de relevancia propuesto mejora estadísticamente la clasificación de tareas MI con un conjunto reducido de canales. Evaluamos ambos métodos en un conjunto de datos bi-clase de motor-imagery (MI) que contiene 64 canales y m\'{a}s de 40 sujetos. El tercer método propone una medida de relevancia basado en la proyección de los datos en un espacio de alta dimensión a partir de método de kernel trick en el cual se usa como medida de similitud la correntropia, esto permite que el método sea robusto a las variaciones temporales y la variabiliad presente en las diferentes pruebas. A partir de esta medida se realiza una selección de patrones que mejora significativamente el desempeño de clasificación en una tarea de motor imagery bi-clase (MI) en una base de datos publica compuesta por 22 canales y nueve sujetos. |
dc.format.extent | 44 |
dc.format.mimetype | application/pdf |
dc.language.iso | eng |
dc.rights | Derechos reservados - Universidad Nacional de Colombia |
dc.rights.uri | http://creativecommons.org/licenses/by-nd/4.0/ |
dc.subject.ddc | Ingeniería y operaciones afines::Otras ramas de la ingeniería |
dc.title | Time series analysis using information measures supporting supervised learning tasks |
dc.title.alternative | Análisis de series de tiempo utilizando medidas de información que apoyan tareas de aprendizaje supervisado |
dc.type | Otro |
dc.rights.spa | Acceso abierto |
dc.type.driver | info:eu-repo/semantics/other |
dc.type.version | info:eu-repo/semantics/acceptedVersion |
dc.contributor.researchgroup | Grupo de Control y Procesamiento Digital de Señales |
dc.description.degreelevel | Doctorado |
dc.publisher.department | Departamento de Ingeniería Eléctrica y Electrónica |
dc.publisher.branch | Universidad Nacional de Colombia - Sede Manizales |
dc.relation.references | [1] AHN, Minkyu ; JUN, Sung C.: Performance variation in motor imagery brain–computer interface: a brief review. In: Journal of neuroscience methods 243 (2015), S. 103–110 [2] AL-ANI, A. ; AL-SUKKER, A.: Effect of Feature and Channel Selection on EEG Classification. In: 2006 International Conference of the IEEE Engineering in Medicine and Biology Society, 2006. – ISSN 1557–170X, S. 2171–2174 [3] AL-ANI, A. ; AL-SUKKER, A.: Effect of Feature and Channel Selection on EEG Classification. In: 2006 International Conference of the IEEE Engineering in Medicine and Biology Society, 2006, S. 2171–2174 [4] ALIMARDani, Fatemeh ; BOOSTANI, Reza ; BLANKERTZ, Benjamin: Weighted spatial based geometric scheme as an efficient algorithm for analyzing single-trial EEGS to improve cue based BCI classification. In: Neural Networks 92 (2017), S. 69–76 [5] A´LVAREZ-MEZA, A.M. ; CA´RDENAS-PEn˜A, D. ; CASTELLANOS-DOM´INGUEZ, G.: Unsupervised Kernel Function Building Using Maximization of Information Potential Va- riability. Cham : Springer International Publishing, 2014. – 335–342 S. – ISBN 978–3– 319–12568–8 [6] A´LVAREZ-MEZA, Andr´es M. ; CA´RDENAS-PEn˜A, David ; CASTELLANOS-DOMINGUEZ, Germ´an: Unsupervised kernel function building using maximization of information potential variability. In: Iberoamerican Congress on Pattern Recognition Springer, 2014, S. 335–342 [7] A´LVAREZ-MEZA, Andr´es M. ; VELA´SQUEZ-MARt´INEZ, Luisa F. ; CASTELLANOS- DOMINGUEZ, Germ´an: Time-series discrimination using feature relevance analysis in motor imagery classification. In: Neurocomputing 151 (2015), S. 122–129 [8] ArvANEH, M. ; GUAN, C. ; ANG, K. K. ; QUEK, C.: Optimizing the Channel Selection and Classification Accuracy in EEG-Based BCI. In: IEEE Transactions on Biomedical Engineering 58 (2011), June, Nr. 6, S. 1865–1873. – ISSN 0018–9294 [9] BRIDWELL, David A. ; RACHAKONDA, Srinivas ; SILVA, Rogers F. ; PEARLSON, God- frey D. ; CALHOUN, Vince D.: Spatiospectral Decomposition of Multi-subject EEG: Evaluating Blind Source Separation Algorithms on Real and Realistic Simulated Data. In: Brain Topography 31 (2018), Jan, Nr. 1, S. 47–61. – ISSN 1573–6792 [10] Cho, Hohyun ; AHN, Minkyu ; AHN, Sangtae ; KwON, Moonyoung ; JUN, Sung C.: EEG datasets for motor imagery brain computer interface. In: GigaScience 6 (2017), 05, Nr. 7. – ISSN 2047–217X [11] DAI, Shengfa ; WEI, Qingguo: Electrode channel selection based on backtracking search optimization in motor imagery brain-computer interfaces. In: Journal of Integrative Neuroscience 16 (2017), feb, Nr. 3, S. 241–254. – ISSN 1757448X [12] DECETY, Jean ; INGVAR, David H.: Brain structures participating in mental simulation of motor behavior: A neuropsychological interpretation. In: Acta psychologica 73 (1990), Nr. 1, S. 13–34 [13] DECETY, Jean ; STEVENS, Jennifer: Action representation and its role in social inter- action. In: The handbook of imagination and mental simulation (2009), S. 3–20 [14] ELASUTY, Basem ; ELDAWLATLy, Seif: Dynamic Bayesian Networks for EEG motor imagery feature extraction. In: 2015 7th International IEEE/EMBS Conference on Neural Engineering (NER) IEEE, 2015, S. 170–173 [15] FRANKLIN ALEX JOSEPH, A. ; GoVINDARAJU, C.: Channel selection using glow swarm optimization and its application in line of sight secure communication. In: Cluster Computing (2017), sep, S. 1–8. – ISSN 15737543 [16] GRETTON, Arthur ; BorgwARDT, Karsten ; RASCH, Malte J. ; SCHOLKOPF, Bernhard ; SMOLA, Alexander J.: A Kernel Method for the Two-Sample Problem. (2008), may [17] GUERRERO-MOSQUERA, C. ; NAVIA-VazquEz, A.: Automatic removal of ocular ar- tefacts using adaptive filtering and independent component analysis for electroencep- halogram data. In: IET Signal Processing 6 (2012), April, Nr. 2, S. 99–106. – ISSN 1751–9675 [18] HANAKAWA, Takashi ; DIMYAN, Michael A. ; HALLETT, Mark: Motor planning, ima- gery, and execution in the distributed motor network: a time-course study with functio- nal MRI. In: Cerebral cortex 18 (2008), Nr. 12, S. 2775–2788 [19] KEE, Chea Y. ; PONNAMBALAM, S. G. ; Loo, Chu K.: Multi-objective genetic algorithm as channel selection method for P300 and motor imagery data set. In: Neurocomputing 161 (2015), aug, S. 120–131. – ISSN 18728286 [20] Krucoff, Max O. ; RAHIMPOUR, Shervin ; SLUTZKY, Marc W. ; EDGERTON, V R. ; TURNER, Dennis A.: Enhancing nervous system recovery through neurobiologics, neural interface training, and neurorehabilitation. In: Frontiers in neuroscience 10 (2016), S. 584 [21] LIANG, Shuang ; Choi, Kup-Sze ; QIN, Jing ; WANG, Qiong ; PANG, Wai-Man ; HENG, Pheng Ann: Discrimination of motor imagery tasks via information flow pattern of brain connectivity. In: Technology and Health Care 24 (2016), Nr. s2, S. S795–S801 [22] LUNA-NARANJO, David ; CA´RDENAS-PEn˜A, David ; CASTELLANOS-DOMINGUEZ, Germ´an: Entropy-Based Relevance Selection of Independent Components Supporting Motor Imagery Tasks. In: International Workshop on Artificial Intelligence and Pattern Recognition Springer, 2018, S. 359–367 [23] LUNA-NARANJO, DF ; HurTADO-RINCON, JV ; CA´RDENAS-PEn˜A, D ; CASTRo, VH ; TORRES, HF ; CASTELLANOS-DOMINGUEZ, G: EEG Channel Relevance Analysis Using Maximum Mean Discrepancy on BCI Systems. In: Iberoamerican Congress on Pattern Recognition Springer, 2018, S. 820–828 [24] MEINICKE, Peter ; RITTER, Helge: Quantizing density estimators. In: Advances in Neural Information Processing Systems, 2002, S. 825–832 [25] MIAo, Minmin ; ZENG, Hong ; WANG, Aimin ; ZHAo, Changsen ; Liu, Feixiang: Dis- criminative spatial-frequency-temporal feature extraction and classification of motor imagery EEG: An sparse regression and Weighted Na¨ıve Bayesian Classifier-based ap- proach. In: Journal of neuroscience methods 278 (2017), S. 13–24 [26] PRINCIPE, Jose C.: Information theoretic learning: Renyi’s entropy and kernel perspec- tives. Springer Science & Business Media, 2010 [27] Qiu, Zhaoyang ; JIN, Jing ; LAM, Hak-Keung ; ZHANG, Yu ; WANG, Xingyu ; CICHOCKI, Andrzej: Improved SFFS method for channel selection in motor imagery based BCI. In: Neurocomputing 207 (2016), S. 519 – 527. – ISSN 0925–2312 [28] SAIOTE, Catarina ; TACCHINO, Andrea ; BRICHETTO, Giampaolo ; ROCCATAGLIAta, Luca ; BOMMARITO, Giulia ; CORDANO, Christian ; BATTAGLIA, Mario ; MANCAR- DI, Giovanni L. ; INGLESE, Matilde: Resting-state functional connectivity and motor imagery brain activation. In: Human brain mapping 37 (2016), Nr. 11, S. 3847–3857 [29] SchO¨LKOPF, Bernhard ; SMOLA, Alexander J. ; BAch, Francis [u. a.]: Learning with kernels: support vector machines, regularization, optimization, and beyond. MIT press, 2002 [30] WOOD, Anthony J. ; BLYTHE, Richard A. ; EvANS, Martin R.: R´enyi entropy of the totally asymmetric exclusion process. In: Journal of Physics A: Mathematical and Theoretical 50 (2017), Nr. 47, S. 475005 [31] YANG, Huijuan ; GUAN, Cuntai ; WANG, Chuan C. ; ANG, Kai K.: Maximum depen- dency and minimum redundancy-based channel selection for motor imagery of walking EEG signal detection. In: ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings, IEEE, may 2013. – ISBN 9781479903566, S. 1187– 1191 [32] Zich, Catharina ; DEBENER, Stefan ; KRANCZIOCH, Cornelia ; BLEICHNER, Martin G. ; GUTBERLET, Ingmar ; DE Vos, Maarten: Real-time EEG feedback during simultaneous EEG–fMRI identifies the cortical signature of motor imagery. In: Neuroimage 114 (2015), S. 438–447 |
dc.rights.accessrights | info:eu-repo/semantics/openAccess |
dc.subject.proposal | Selección de Componentes |
dc.subject.proposal | Component selection |
dc.subject.proposal | Entropía de Renyi |
dc.subject.proposal | Renyi Entropy |
dc.subject.proposal | Channel selection |
dc.subject.proposal | Análisis de relevancia de series de tiempo |
dc.subject.proposal | Interfaz cerebro máquina |
dc.subject.proposal | Time-series Relevance analysis |
dc.subject.proposal | Brain Computer Interface |
dc.type.coar | http://purl.org/coar/resource_type/c_1843 |
dc.type.coarversion | http://purl.org/coar/version/c_ab4af688f83e57aa |
dc.type.content | Text |
oaire.accessrights | http://purl.org/coar/access_right/c_abf2 |
Archivos en el documento
Este documento aparece en la(s) siguiente(s) colección(ones)
![Atribución-SinDerivadas 4.0 Internacional](/themes/Mirage2//images/creativecommons/cc-generic.png)