Research On Human Movement Intention Recognition Based On Motor Imagery EEG Signals

The exoskeleton robot is a special existence in the field of robotics,which is a system where humans and machines coexist.It is attached to the human body and cannot act independently,assisting the human body to perform various tasks.When the human-computer system uses the EEG signal as the signal source,it can control the action of the exoskeleton by analyzing the EEG signal to identify the human movement intention,so that a typical brain-computer interface(Brain-Computer Interface,BCI)system can be constructed.With the help of this exoskeleton robot as an actuator brain-computer interface system,rehabilitation medical training is carried out for patients who have lost their ability to exercise due to brain injury.It rebuilds the patient’s motor neural circuits,thereby improving the mobility of the limbs.This article first introduces the structure of the brain and the functional division of various brain regions,so as to determine the region where the left and right motor cortex is located.From a microscopic perspective,the electrical activity of a single neuron is described,and then to a macroscopic perspective,countless neurons work synchronously to generate EEG signals.EEG signals are non-stationary and random,which is not conducive to conventional time-domain or frequency-domain analysis.Studies have shown that time-frequency analysis is very effective for EEG signals.EEG signals are composed of waves of different frequency bands,each of which appears on a specific person or scene and has its own characteristics.After introducing the frequency band characteristics of EEG,it is convenient to elicit the ERD/ERS phenomenon related to mu rhythm and beta wave,so as to provide a clear direction for subsequent preprocessing and feature selection.Finally,the method of EEG acquisition and the problems to be paid attention to are introduced.The data source of this paper is the data set Data set III released by the BCI competition in 2003,which recorded the EEG data of 140 groups of left and right motor images of a subject.The ERD/ERS phenomenon occurs in the mu rhythm and beta wave,that is,in the frequency range of 8～30Hz.Therefore,the original signal with a frequency range of 0.5 to 30 Hz is subjected to band-pass filtering processing of 8 to 30 Hz,so as to intercept the motor imagery EEG signal.From the perspective of time-frequency-space,the features of different domains are extracted.The Hjorth parameter is used to extract the time-domain features,the wavelet packet decomposition is used to extract the time-frequency features-wavelet packet energy and relative wavelet packet energy,and the co-spatial mode is used to extract the spatial domain features.Finally,the parameters of the feature extraction algorithm are selected,and the extracted features are analyzed.All features are fused to construct fused features.Due to the insufficient sample size of the public data set selected in this paper,after trying to use the convolutional neural network to conduct motor imagery EEG classification research,it ended with less than ideal classification results.Thus returning to traditional machine learning algorithm research to tap its potential.Firstly,the research on human motion intention recognition is carried out through two traditional single model machine learning algorithms-decision tree and support vector machine.The ensemble learning algorithm XGBoost is introduced to build an ensemble model based on a decision tree.A study of EEG classification based on multiple fusion features.Finally,in order to further improve the accuracy of EEG classification,the decision tree,support vector machine and XGBoost are fused by stacking method.The expected effect is obtained.This lays a solid theoretical foundation for the subsequent practical application of brain-computer interface technology to the control of exoskeleton robots.