Research On Facial Expression Decoding Based On FMRI Brain Signal

Medical image processing is an important branch of image recognition technology.it has also received widespread attention.Along with the rapid development of deep learning technology in recent years,the accuracy of image classification by various machine learning algorithms has been continuously improved.Therefore,image recognition technology has also received extensive attention.Image analysis technology and classifier performance are two important research directions in the field of image recognition.With the rapid development of image recognition technology,more and more people are applying image recognition technology and machine learning algorithms to the processing of medical images to use artificial intelligence to analyze complex medical images and find out the hidden mysteries in medical images.In the study of medical image recognition,the study focused on fMRI images of the DICOM format obtained by nuclear magnetic resonance.In the research field of fMRI images,the analysis of brain images has become a hot issue in the study of fMRI images.In terms of brain image analysis,research and analysis of human emotions are the main research directions.If we can understand the facial expressions of the experimenters participating in the experiment by analyzing the fMRI images,then we can better understand the brain and understand the behavior of the human body controlled by each brain partition.According to relevant literature,there are six types of human facial expressions.They are "anger","disgust","sadness","happiness","joy",and "surprise".This paper will use two kind facial expressions("anger","joy")to parse.In this paper,the brain image of the subject after stimulation is obtained by using a nuclear magnetic resonance apparatus.At the same time,the corresponding facial expression is found through the time series.The obtained brain image is subjected to preprocessing and SPM analysis,and then the feature vector is obtained.Because the dimension of the obtained feature vector is higher,input classifier directly to classify will affect the result of the experiment.Therefore,this experiment first uses dimension reduction technique principal component analysis(PCA)to reduce the dimension of the feature vector,and then obtains the obtained feature vector input classifier extreme learning machine uses a cross-validation method to count the accuracy of the parsed expression.In order to better compare the performance of the classifier,this paper applies the two algorithms of support vector machine and extreme learning machine to the same data set.The experimental results show that the extreme learning machine algorithm model results are superior to the support vector machine algorithm model.