Research On Brain Computer Interface Technology Based On Hilbert Huang Transform And Support Vector Machine Optimization

Brain computer interface technology(BCI)is a kind of information communication mode that can realize the direct connection between brain and computer without relying on the conventional information transmission path of brain.Brain computer interface provides an opportunity for patients with damaged brain neural pathways to communicate with the outside world.In addition to rehabilitation field,BCI also has important applications in military,artificial intelligence,remote control and entertainment fields.Among them,the analysis of motor imagery EEG(MIEEG)is the focus of BCI research.Therefore,it is of great significance to build a motor imagery brain computer interface system(MIBCI)to analyze and process MIEEG signals in scientific research and practical applications.This paper first introduces the background and application of BCI.The data is preprocessed and the features of MIEEG are extracted by Hilbert Huang transform(HHT).The support vector machine(SVM)is used for pattern recognition of the extracted EEG features,and then the corresponding parameters are optimized by genetic algorithm(GA),particle swarm optimization(PSO)and artificial bee colony algorithm(ABC).Based on the off-line data analysis coming from BCI competitions,the two classification and four classification MIEEG are tested respectively.Finally,the MIBCI platform is built and the off-line data is applied to the system.The main contents of this paper are as follows:Firstly,this paper summarizes the related knowledge of EEG and motor imagery EEG,and introduces the background,development and application of BCI.Secondly,this paper studies the feature extraction method in BCI.Several common feature extraction methods are introduced and the feature extraction method of EEG based on Hilbert Huang transform is adopted.This method is self-adaptive and can process non-stationary and non-linear signals.The marginal spectrum obtained by Hilbert Huang transform can reflect the relationship between energy and frequency,which plays an important role in signal analysis and can analyze the variation of signal amplitude with frequency.Then,the paper discusses the commonly used classification methods of EEG signals,and analyzes the advantages of SVM method in the classification of extracted EEG features,thus SVM is more suitable for the data processing module of brain computer interface.When the samples are linearly separable and nonlinearly separable,the classification process of SVM for two and four kinds of samples is discussed respectively.In order to get better classification accuracy,this paper proposes a method to classify the extracted EEG features using SVM optimized by GA,PSO and ABC algorithm.Experimental analysis shows that SVM optimized by ABC algorithm can get better classification effect.Two datasets are used in two experiments,which are from datasets ? in BCI competition ? and datasets 2a in BCI competition ?.Datasets ? stands for two classification(left hand,right hand)and datasets 2a stands for four classification(left hand,right hand,tongue,feet)datasets respectively.In this paper,two datasets are tested based on the above methods.In the experiment,the feature of EEG is extracted by HHT,and then SVM is used for classification.Finally,the parameters of GA,PSO and ABC algorithms are optimized.Among them,the classification accuracy of experiment 1 is up to 100%,and that of experiment 2 is up to 86.1111%.Finally,based on the above feature extraction and pattern recognition methods,an online MIBCI system is built by using MATLAB and Python.During the system,the GUI control interface is designed through Qt Designer.The acquisition and data processing program written by MATLAB provides corresponding functions to the control interface.Finally,the control interface is converted into a Python program and packaged into an executable file.The system effectively simulates the control process of one small ball,and the small ball represents external equipment.The experimental results show that the system has good performance.