Epilepsy Signal Detection Algorithm Based On Wavelet Transform

Epilepsy is a common chronic brain neurological disease.The cause of seizures is sudden abnormal discharge of neuronal cells in the brain.It is the second most common ailment in neurological diseases after cerebrovascular disease.Epilepsy usually has the following abnormal symptoms during the episode: sputum,confusion,intermittent,stiff,short-lived motion,sudden syncope,limb twitching,and foaming at the mouth.Studying epilepsy signals is of great significance for diagnosing epilepsy,locating epileptic foci,distinguishing the type of epileptic seizures,and reducing the frequency of epileptic seizures and the incidence of seizure disability and lethality.After related research,this paper proposes an algorithm based on wavelet transform for epilepsy signal detection.The algorithm is divided into three parts: epileptic signal preprocessing,feature extraction and epilepsy signal detection.The algorithm uses the improved wavelet threshold function to remove the artifact signal in the epileptic signal;The features of epileptic EEG signals were extracted by wavelet multiresolution analysis and nonlinear dynamics correlation algorithm.And using the nonlinear classifier in the support vector machine to detect and classify the epileptic signals.The specific research work of the algorithm is divided into the following three aspects:Firstly,the wavelet threshold is studied and an improved wavelet threshold function is constructed for denoising of epileptic signals.According to the similarity between the characteristic wave of the epileptic signal and the wavelet base,the appropriate wavelet basis function and the decomposition scale are selected to make the discrete wavelet transform of the epileptic signal.Then extracts the detail coefficients and do the threshold quantization process,and finally reconstructs the epilepsy signal according to the wavelet coefficients.By analyzing the signal-to-noise ratio and the minimum mean square error before and after denoising,the improved signal-to-noise ratio of the threshold function is larger and the root mean square error is smaller,and the signal-to-noise ratio is improved by at least 21% compared with the soft and hard threshold functions.The root mean square error is correspondingly reduced by 35%.Secondly,the non-stationary characteristics of epileptic EEG signals were selected by nonlinear dynamics algorithm for feature extraction of epileptic signals.The paper proposes an epileptic signal feature extraction algorithm based on wavelet transform multi-resolution analysis,Teager energy operator and sample entropy.By extracting the signals in the specific frequency range by wavelet transform of the epileptic signal.After the Teager energy operator is processed,the sample entropy analysis is performed to obtain the nonlinear characteristics of the epileptic signal.By analyzing the difference of sample entropy between normal EEG signals and epileptic signals,this feature extraction algorithm is effective and feasible.Finally,the support vector machine is used to classify the epilepsy signals.The training set and the test set are selected according to the extracted feature vector,and the feature vector is classified by the nonlinear classifier in the support vector machine.For the penalty parameters and kernel function parameters,the paper uses the genetic algorithm to optimize the parameters.The simulation results show that the training set of normal EEG and epilepsy signals has the highest classification accuracy,reaching 96.667%,and the classification accuracy of epilepsy signal is 91.667%.The classification of epileptic signal detection is the best.