Research on prediction of epileptic EEG signal based on Randomly-Distributed-Embedding Model

Epilepsy is a common neurological disease characterized by abnormal electrical discharges in the brain. Electroencephal ogram (EEG) is widely used to diagnose possible epileptic seizures. Many researchers have been devoted to predicting the seizures and abnormal discharges of brain regions by analyzing the nonlinear characteristics of EEG signals. Predicting the future state of a nonlinear dynamic system such as EEG signals is a difficult task, especially when only short-term and high-dimensional EEG signal samples are available in the real system. Therefore, this paper proposes a method based on the Randomly Distributed Embedding (RDE) method. We first use the model sequence generated by the Rössler system to analyze the effectiveness of the RDE model for predicting nonlinear dynamic systems, and then apply the RDE model to the prediction of the EEG signals of normal people and epilepsy patients. Under the harsh premise of small amount of data with high dimensions, it overcomes the shortcomings of traditional machine learning prediction algorithms that require subst antial training data volume. It converts the high-dimensional hindrance to prediction into a source of practical information, which can accurately predict the future state of EEG signals of normal people and epilepsy patients.