Research On EEG Emotion Recognition Algorithm Based On FPGA

Emotion is human’s subjective cognitive experience.EEG emotion recognition refers to the process that artificial intelligence automatically recognizes the emotional state by acquiring individual EEG signals.Aiming at the problems of dimensional disaster caused by too many extracted features and resulting in low classification accuracy in the traditional EEG emotion recognition process.An EEG emotion recognition model based on the BAS-SVM optimized feature selection is proposed in this thesis.FPGA is able to process EEG data in parallel and can realize the porting of emotion recognition algorithm,so this thesis conducts the research of EEG emotion recognition algorithm based on FPGA.Firstly,the EEG data of the DEAP dataset were divided into four frequency bands,and the alpha wave,beta wave and theta wave rhythm signals of EEG were decomposed and reconstructed by the Mallat algorithm,and the sample entropy,energy and power spectral density were extracted as EEG features.Secondly,the results of different swarm intelligence algorithms are compared to optimize EEG feature selection.In this thesis,the beetle antenna search(BAS)algorithm is used to discretize and binarize the characteristic matrix.The feature subset is introduced into the objective function to search the optimal feature subset.Finally,support vector machine is used to classify EEG emotion.The experimental results show that the classification accuracy of the BAS-SVM algorithm proposed in this thesis is improved by 10.99% and 7.07% compared with the GA-SVM and the PSO-SVM,respectively.Compared with the traditional dimensionality reduction method,the data dimensionality is reduced to 11 dimensions,eliminating redundant features and avoiding dimensional disasters,and the EEG emotion recognition accuracy is90.72%,with an accuracy improvement of about 10%.With the miniaturization of devices and the popularity of the concept of electronic consumption,EEG signal processing algorithms need to be ported to hardware devices.In the FPGA design of EEG emotion recognition algorithm,according to the principle and flow of the algorithm,the hardware programming language is used to realize the modular design of the algorithm,and the wavelet transform module,timing control module,output valid bit module,output selection module,and test module are designed,and the functions and simulation results of each module are elaborated.Firstly,the EEG signal is passed through the wavelet four-layer decomposition module for EEG frequency band division to obtain 54-dimensional EEG features.Secondly,the product of the projection matrix of the11-dimensional optimal feature subset obtained by the BAS-SVM and the optimal weight coefficient vector is solidified into ROM.Finally,the projection matrix is multiplied with the optimal weight coefficient vector of the SVM,and the product result is summed with the bias to obtain the output result,which achieves the classification of EEG emotional signals.The simulation results show that the Kappa coefficient is 0.8144,and the EEG emotion recognition model has certain stability and practicality.