Research On Analysis Model And Algorithm Of Time-varying Signal Based On Deep Neural Network

The problem of nonlinear time-varying signals involves many fields such as medical,audio,economic,and oilfield mining.Currently,the classification and prediction of time-varying signal data is one of the current research hotspots in the field of signal processing and artificial intelligence,which has strong theoretical and practical significance.Due to the wide application of time-varying signals,and the characteristics of time-varying signals,such as high time-dependence,non-stationary,high-dimensional,noise,variable modal characteristics,irregular,time-varying signal classification has been one of the difficulties in the field of classification.With the development of technology and the improvement of the performance of hardware facilities,deep neural network has gradually replaced the traditional machine learning algorithm in computer vision,signal and natural language processing and become the mainstream of the industry,and achieved good results.Compared with the traditional neural network,the process neural network is an extension of its time domain.In mechanism,it is the accumulation of time effect and can express the common effect of many factors at the same time.It has its own unique advantages in dealing with time-related problems.Therefore,if the deep learning method is combined with the process neural network,it will produce better complementary effects in the field of time-varying signals.Therefore,if the deep learning method is combined with the process neural network,it will produce better complementary effects in the field of time-varying signals.In this paper,the process neural network is introduced into the classification of time-varying signals,and combined with related technologies such as denoising autoencoder,the research on the classification of arrhythmia based on ECG signals in the biomedical background is carried out.The work and innovation of this paper include the following aspects:1.The feature extraction methods of time-varying signals are studied,and the process neural network and self-encoder are elaborated.2.Aiming at the classification of cardiac arrhythmias in cardiovascular diseases,taking single heart beat signal as the research object,a denoising autoencoder deep process neural network(DAE-DPNN)based on orthogonal basis of trigonometric function is proposed.By using the algorithm strategy based on orthogonal function basis expansion,the implicit expression of input time-varying signal is realized.Experimental results show that the model and algorithm are effective and feasible.3.Aiming at the problem of complex long-signal classification and prediction,based on the idea of convolutional neural network feature extraction,a denoising autoencoder deep convolution process neural network(DAE-DCPNN)is proposed.Taking 12-lead ECG signal as research object,the diagnosis and prediction of arrhythmia were achieved,and the validity of the model was verified;the original unenhanced data and the enhanced data were used for comparative experiments,and the degree of influence of data balance on the model was discussed.In this paper,for different application scenarios of time-varying electrocardiogram signals in biomedicine,two different process neural network combination models are designed and implemented and related experiments are carried out,and good results are obtained.