Recognition And Classification Of Motor Imagery EEG Signals Based On Probabilistic Collaborative Representation

Exploring the working mechanism of the brain has always been an important challenge for scholars in the field of neuroscience.The brain completes communication and information exchange with the external environment through peripheral nerves and muscle channels.In the real world,many people suffer severe motor disorders and lose the basic ability to communicate with the outside world.The brain computer interface(BCI)does not rely on the normal output channel of the brain and establishes a direct channel for information transmission between the brain and the outside world.The BCI system,which converts different brain state of mind into different control signals through the analysis and processing of the collected brain signals,so as to control external equipment.As a rising human-computer interaction pattern,BCI provides a new communication and control channel for patients with severe dyskinesia.At present,the brain computer interface has become a research focuses in many spheres such as neuroscience,computer communication and control,biomedical engineering and so on.The brain signals that evoked by the change of brain motor cortex rhythm through motor imagery,which is considered as a theoretical basis for the study of motor imagery based BCI system.Signal processing part including preprocessing,feature extraction and classification as the core link,which has great influence on the property of the whole BCI system.This paper is based on the research of motor imagery based BCI system,and studies the feature extraction and classification recognition of brain signals.A method of brain signals recognition based on S-transform(ST)and probabilistic collaborative representation based classification is proposed.The feasibility of the proposed algorithm is verified by the international standard BCI contest database and collected brian signals.The main research content and innovation of this article are as bellow:Firstly,this paper briefly presents the context and significance of the BCI system,outlines the definition and composition of the BCI system,and elaborates the research status at home and abroad,and summarizes the problems and challenges faced by BCI technology.This paper also briefly introduces brain signals,invasive system and non invasive system,mainly expatiates the theoretical principle of motor imagery.Secondly,the algorithm of feature extraction is studied in this paper,the quality of feature extraction will directly affect the effect of the rear classification.As a stretch of short time Fourier transform and wavelet transform,S-transform has the ability of multi-resolution analysis of the change of window function with frequency,and it is same with the handling of non-stationary brain signals.In this paper,we use S-transform to extract features of signals and select power spectral density after S-transform as features,so we can accurately locate the spectrum change of sensorimotor rhythm.Then,the algorithm of classification algorithm is studied,and this paper proposes a probabilistic collaborative representation based classifier(ProCRC)and robust ProCRC(R-ProCRC)to realize the recognition of two types of motor imagery signals.The main idea of the algorithm is jointly maximizes the likelihood that a test sample belongs to each of the multiple classes.The ultimate classification is implemented by checking which class has the maximum likelihood,and which has a clear probabilistic interpretation.Finally,the experimental design procedure of motor imagery was introduced in detail,and the acquired electroencephalogram(EEG)signals and electrocorticography(ECoG)signals from database were used as experimental data.The experimental results show that compared with the traditional classification algorithm,such as bayesian linear discriminant analysis and gradient Boosting,the features extracted from the S transform and the classification algorithm based on the probabilistic collaborative representation can get better classification results and better timeliness.It provides a good reference value for the future application of brain signals recognition.In this paper,we first applied the probabilistic collaborative representation based classification algorithm to the recognition of brain signals,and achieved good results.