Particle Swarm Optimization Based Support Vector Machines in Motion Imagery

In order to improve the classification recognition rate, this paper carries out a theoretical analysis of the support vector machine (SVM), in order to find the optimal parameters C, g, the particle swarm algorithm optimized support vector machine (PSO-SVM) is introduced to carry out parameter optimization, and for the shortcomings of the PSO-SVM, the optimal classifier, i.e., Adaptive Particle Swarm Algorithm Optimized Support Vector Machine (APSO-SVM), is introduced and the extracted features are respectively inputted into the extracted features are fed into the optimal classifier APSO-SVM respectively. The final classification results show that the proposed feature fusion analysis and the introduced APSO-SVM can effectively classify MI EEG signals.

I. Introduction

At present, the BCI paradigm based on motor imagery EEG signals, i.e., MI-BCI system, is one of the most popular paradigms nowadays ^[1] , which has not only been successfully applied in the recovery of patients with physical disabilities and movement disorders, but also achieved very remarkable results in the fields of neuroscience, brain cognition, and aerospace science and technology. In recent years, due to various reasons, the proportion of patients with physical disabilities and acromegaly in the population is increasing year by year, and with the gradual arrival of the aging society, the elderly has been a large group that can not be ignored ^[2] . With the gradual arrival of an aging society, the elderly have become a large group that cannot be ignored. In the next two decades, the number of elderly people over the age of 60 will double ^[3] , and the pressure on society as a whole will increase dramatically. They share common characteristics: limited mobility, poor self-care ability, poor quality of life and sense of well-being, and the need to be cared for in their lives. The emergence of MI-BCI technology provides a novel way for such people to achieve self-care and communicate with the outside world. In this paper, through the study of lower limb movement imagery ^[4] , the classification prediction model based on particle swarm optimization support vector machine is established by using particle swarm optimization algorithm to optimize the optimal parameter ^[5] , which is difficult to find in the traditional support vector machine model. The experimental results show that compared with the traditional support vector machine model, the optimized model prediction accuracy is improved by 7.8%~11.5%, and the model training time is also shortened.