Emotion Recognition of EEG Signals Based on Improved GAF Algorithm
Human emotion recognition using electroencephalography(EEG)signals is an important and challenging research area.The traditional method is to analyze the one-dimensional EEG signal,and then extract features for identification.But this method needs to extract many features in time domain and frequency domain to achieve better identification effect.Considering that the information contained in two-dimensional images is much richer than that contained in one-dimensional signals,converting one-dimensional signals into two-dimensional images can extract more effective features for recognition.We propose an EEG emotion recognition method based on an improved Gramian Angular Field(GAF)algorithm.First,the sub-band signals of alpha,beta,and gama are extracted from the EEG signal.Then,an improved GAF algorithm based on Mahalanobis distance weighting is proposed to convert the one-dimensional EEG signal into a two-dimensional featured images,and features such as singular value entropy and graph energy are extracted from the two-dimensional featured images.Finally,convolutional neural network(CNN)is used to classify the extracted EEG features.Based on the widely used DEAP dataset,the model is validated for the four-class(HAHV,LAHV,LALV,and HALV)emotion recognition task.The experimental results show that the average classification accuracy of the proposed algorithm reaches 92.63%,which has certain advantages compared with the existing recognition methods.
国家科技期刊平台
NSTL
万方数据
