Dynamic Pattern Recognition Based BCI System And Rehabilitation Application

In recent years,the number of stroke patients shows increasing tendency.Brain damage and movement disorder caused by stroke put tremendous pressure on both patients and their families.Traditional clinical treatment combined with brain computer interface system training,as a new way of rehabilitation,can utilize patients' intention effectively and enhance the efficiency of the brain motor function rehabilitation.When applying the BCI based system in training,we found there are some differences between signals of stroke patients and the normals in performing motor imaginary.To this end,we employ dynamic pattern recognition method to improve the recognition rate of stroke patients' signal based on motor imagery signal of normal people.We introduce a BCI rehabilitation system in detail and prove its effectiveness by the clinical experiments we operate.For such a motor imagery based BCI system,it is important to recognize patient's imagine pattern in high accuracy,which directly affect the feasibility of rehabilitation system.Since we cannot directly apply traditional methods to EEG of stroke patients,we combine tensor based feature and deep network together to deal with EEG of stroke patients and compare different dynamic feature extraction methods and pattern recognition methods.We analyze signals of stroke patients from effect of rehabilitation training,online and offline data analysis and conduct experiments on dataset of normal people.We also analyze the difference between signals of stroke patients and the normals,compare the results from the experiment group and the control group.We further investigate the dynamic nature of EEG in the rehabilitation process.The results not only demonstrate our method improve recognition performance on motor imagery signals of stroke patients,but also evidence the feasibility of our BCI system on clinical rehabilitation training.