Research On Human Physiological Signal Classification Based On Genetic Algorithm And Multilayer Perceptron

Human physiological signal is the embodiment of human life information,and is a window of life phenomenon.The physiological signals,such as electromyography and electroencephalogram,are the main indexes to reflect the physical activity of the human body and contain a lot of physiological information,for example,EEG is a very important tool in the diagnosis of epilepsy.Timely discovery of their changes can help people to prevent diseases in advance and reduce the occurrence of some unexpected situations,which is of great significance.This thesis mainly studies the classification of two signals(electromyography and electroencephalogram)of the human body.The human physiological signals is a chaotic signal,which has the characteristics of nonlinear dynamics,and it also has the characteristics of randomness and nonstationarity.According to these characteristics,we extracted the features using time domain,frequency domain,empirical mode decomposition,and a genetic algorithm-based feature searching method;then,we performed the optimization selection of the extracted features;finally,based on the selected features,the signals were classified through a MLP classifier.In the experiment,we set the analysis of EEG and EMG signals of two publicly available datasets,the accuracy rate of two datasets were respectively 90.12%and 97.67%and the proposed method is better than the existing classification method in the classification accuracy rate.At the same time,the effectiveness of the method in the classification of human physiological signals is also demonstrated by cross validation.In addition,a window cut preprocessing is also done for EEG datasets.The experimental results show that this method can identify epileptic EEG rapidly,and has broad prospects for clinical application in the future.