Designing a model to detect beta burst in EEG using nonlinear dynamic features based on machine learning