Activity Number:
|
522
- Contributed Poster Presentations: Biometrics Section
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Biometrics Section
|
Abstract #304225
|
|
Title:
|
Bivariate Nonlinear Gaussian Processes with Applications to Brain Signals
|
Author(s):
|
Guillermo Granados Garcia* and Hernando Ombao and Wagner Barreto-Souza
|
Companies:
|
King Abdullah University of Science and Technology and King Abdullah University of Science and Technology (KAUST) and Universidade Federal de Minas Gerais
|
Keywords:
|
nonlinear time series;
Brain Signals Analysis;
Gaussian Processes
|
Abstract:
|
This work is motivated by the need to capture complex possibly non-linear temporal dynamics between components of a bivariate time series. Here, we study the statistical and probabilistic properties of a bivariate Gaussian nonlinear autoregressive time series model. Our first task will be to present the conditions by which stationarity and ergodicity of the process is ensured. We then derive the maximum likelihood estimators of the parameters and investigate the validity of the proposed model by simulating different nonlinear processes studied in the statistical literature and fitting them with our proposal using optimization methods. We use the proposed model over an EEG time series experiment to analyze lead-lag relationship and complex non-linear dependence between activity at a pair of brain regions.
|
Authors who are presenting talks have a * after their name.