| With the continuous development of society,the phenomenon of aging population in the world is becoming more and more obvious,and the elderly are inevitably faced with movement disorders and cardiovascular and cerebrovascular diseases.In addition,air pollution and fast-paced life brought about by economic development have also caused more and more young people to suffer from cardiovascular and cerebrovascular diseases.It is of great significance for their families and the whole society to help patients return to normal life.At present,as a promising technology,brain-computer interface system has been widely used in rehabilitation training,and many gratifying results have been achieved.However,most of the brain-computer interface studies will fix and restrict the model in the experimental design,especially in the movement design of patient rehabilitation,which is far from the practical application.Researchers tend to focus on the changes from rest to exercise,while few people pay attention to whether and when the phased goals in the exercise task are achieved.Therefore,in this subject,the brain-computer interface technology based on functional near infrared spectroscopy is used to monitor the brain blood oxygen information of the subjects when completing the action goal in real time,and to judge whether to achieve the phased goal of the task according to the brain information of the subjects.Real-time adjust the driving force of auxiliary equipment(such as manipulators or rubber gloves)according to the state,so that the patient's own strength combined with the auxiliary force of the equipment can achieve the best clamping strength.The intelligent rehabilitation training of brain-computer control closed loop and hand function is realized based on brain information.In addition,in order to fill the gap in the previous research,the limitations in the experimental design are reduced accordingly,and only the ultimate goal of the action task is given.The main research contents and methods of this paper are as follows:(1)During the experiment,47 subjects were asked to hold the table tennis ball with chopsticks without limiting the clamping strength,and the table tennis ball was taken as the stage goal.In the course of the experiment,the fNIRS equipment was used to record the hemoglobin information of the subjects' brain.Here,the difficulty of the task is increased by the way of clamping the ball to simulate the difficulty of the patient gripping the object.By analyzing the process of clamping the ball and the blood oxygen information after clamping the ball,we can dynamically identify whether the subjects reach the stage goal of the task.(2)In the aspect of clamping intention recognition,in order to improve the rapidity of the algorithm,we calculate the Teager-Kaiser energy operator as the feature.Considering that the sample size is too large and unbalanced,the decision model of random forest(Random Forest)algorithm is adopted in the modeling task.The data of 37 subjects were randomly selected as cross-validation samples,and the other 10 subjects were used as the test set of the model.Using the random forest model,the average recognition rate of the cross-validation set is 99.04%,and the average recognition rate of the test set is 95.00%.(3)The identification method of whether the phased goal of the task is achieved or not is proposed.In order to ensure the timeliness of data processing,the sliding window processing method is adopted.In total,six commonly used time domain features and the correlation features between channels are calculated as the features of the model.At the same time,two different binary classification algorithms of Support Vector Machine and Long Short-Term Memory network are adopted to model,select the optimal model,and further optimize the parameters.Finally,using the GA-SVM model,the average recognition rate of the cross-validation set is 94.76%,and the average recognition rate of the test set is 85.83%.(4)In order to verify the feasibility of the algorithm,a platform is built to meet the verification requirements of clamping action intention and reaching standard intention,and the experimental verification in X and Y direction is realized.Finally,the recognition rates of clamping intention and stage state recognition of online verification are 92.34%and 73.26%,respectively.The results show that it is feasible to use fNIRS to distinguish the phased targets of sports tasks on-line.Based on fNIRS technology,this paper explores the changes of the blood oxygen signal of the brain when the human body achieves the phased goal of the task,and establishes the clamping intention model and the phased target state recognition model.the dynamic discrimination of the phased target of the clamping task is realized.These works can provide theoretical basis and technical prototype for the future research and application of BCI rehabilitation,and aim to promote the development of BCI to a new application direction. |
PDF Full Text Request