Research On Classification Of Depression Based On Deep Learning And Signal Processing

Depression is a kind of mental disorder with persistent depression,and has become one of the most prominent mental diseases in the world.Therefore,it is urgent to find a way to diagnose depression quickly and accurately.In recent years,with the rapid development of science and technology,artificial intelligence methods such as machine learning and deep learning have become a popular direction in the field of data analysis.More and more researchers begin to combine artificial intelligence methods with medical data,explore a new model of mining medical data,and make rapid progress in medical diagnosis technology.In this paper,the EEG signal sample data of 148 patients with depression from a hospital in Sichuan Province were used to study the denoising processing of EEG signals of patients with depression based on signal processing method,and a binary deep learning fusion model was established for patients with single and bipolar depression.The main work of this paper is as follows:Firstly,the original EEG signal data is simulated,and several different signal processing methods are used to de-noise it,the effect of different methods is compared,and the best method and parameters are selected to denoise the real EEG signal.Secondly,the time window method was used to slice EEG data and increase the number of samples for data enhancement,so as to better train the model.Then,six kinds of time series deep learning networks,including one-dimensional convolutional neural network,residual neural network based on two-dimensional convolutional neural network,long and short term memory network,bidirectional long short term memory network and gated recurrent unit network,and CNN-LSTM combined with convolutional neural network and recurrent neural network,were used to establish the model.The original data and the EEG signals after noise cancellation were classified respectively,and the high accuracy of more than 90%was obtained on all six models.Finally,D-S evidence theory and majority voting method were used to make decision fusion of the 6 models.The overall highest accuracy rate was 98.56%in the classification of the two types of patients,97.78%in monophasic patients and 99.37%in bipolar patients.It shows that the model established in this paper has excellent classification effect on the EEG data of patients with bipolar depression,and provides a reference for doctors in the clinical diagnosis of depression.