Sensitive text information sensing brain response detection model based on dynamic network
Aiming at the uncerainty of brain response latency caused by complexity of text sensitive information perception process and individual differences,a brain response detection model,based on dynamic convolution neural network with convolutional block attention module(DyCNN_CBAM)is proposed.By adding dynamic convolution module in the model,the convolution parameters of each layer can be changed with the input during training,which can improve the size and capacity of the model.Then,the attention mechanism module is added after the first and second layers of the model to automatically calculate the time-space information with high contribution.Experimental results show that the model improves the average classification accuracy by 4％,and the F1 score by 6.7％,compared with the existing single-scale model;and improves the average classification accuracy by 2％and the F1 score by 1.2％,compared with the existing multi-scale network with simplified network structure.In addition,the best F1 score is obtained on the public dataset.It shows that the network is more suitable for the latency jitter of brain signal sensing by sensitive text information,and effectively improves the stability of sensitive text information detecting model.
sensitive text informationelectroencephalography(EEG)signaltarget detectiondynamic convolutional neural network(DyCNN)attention mechanism
万方数据
维普
