Transfer entropy is an effective method for quantifying information flow and causal orientation in nonlinear systems. Based on this, we combine noise-assisted multivariate variational mode decomposition (NA-MVMD) and transfer entropy together, while considering the case of multivariate systems, and propose kernel based multiscale partial Renyi transfer entropy. We validate the method with henon mapping and VAR model, demonstrating the robustness of the model to noise and the effect of the parameter α taken on the experiment. For the real dataset, we used multi-channel EEG signals performing a mental arithmetic task and computed the information flow of each EEG in different states at different wavebands and compared them separately. In conclusion, the method is able to measure information transfer at different scales, is robust to noise and endpoint effects, and overcomes the interference of other signals in a multivariate system on the target signal, providing a new perspective for exploring the multilevel Spatio-temporal properties of EEG.
Read full abstract