Research On Epileptic EEG System Based On Time-Frequency Analysis

Epilepsy is a brain functional and neurological disorder caused by abnormal discharge of brain neurons.Epilepsy is recurrent and difficult to cure,which affects the life and work of patients seriously,and brings great inconvenience to the family members.At present,EEG is a critical tool for studying seizures.It has high temporal resolution and sensitivity and specificity for the diagnosis of epilepsy.It is an important basis and content for the diagnosis of seizure type and preoperative evaluation of epilepsy.However,clinically,EEG monitoring for patients with epilepsy often lasts for several days.Doctors need to mark and analyze seizures on EEG,which not only brings a heavy workload,but also affects the doctor's judgment.In the paper,the epileptic EEG time-frequency analysis system is designed to collect epileptic EEG signals and extract the time-frequency features.The method based on the BP neural network optimized by improved genetic algorithm is proposed to applied to diagnose epilepsy efficiently.Main content includes:Firstly,a 16-channel epileptic EEG acquisition system was designed using the ADS1299 analog front end to realize the acquisition and conditioning of epileptic EEG signals.Secondly,according to the time-frequency characteristics of epileptic EEG signals,time-frequency analysis of epileptic EEG signals is performed to extract features including the amplitude and power spectrum of epileptic EEG signals,and the relative energy and amplitude of the intrinsic mode function after empirical mode decomposition.Then,the Gaussian mixture model is used to cluster the time-frequency features of epileptic EEG signals.To improve the genetic algorithm,the Expectation-Maximization algorithm is used to estimate the Gaussian mixture model parameters to obtain the optimal parameter combination of the selection operator in genetic algorithm.In order to reduce the BP neural network is too sensitive to the initial weight and threshold,the improved genetic algorithm is used to optimize the BP neural network to obtain the optimal the initial weight and threshold of the neural network,and establish optimized BP neural network classification model.Finally,the model is used to classify and identify epileptic EEG signals to achieve seizure detection.Compared with the BP neural network optimized by traditional genetic algorithm,the system improved the convergence speed of the population and reduced the BP neural network classification error.In the process of automatic diagnosis of epilepsy,the accuracy has also increased by 8.9%,and the classification time reduced by 79.1 seconds.It has important research significance in the diagnosis of clinical seizures.