Semi-supervised Learning Algorithm Is Applied Research In The Brain - Computer Interface

Brain-computer interface (BCI) which translates human's thinking activity to external control signals provides a new communication tool for the patients suffering from peripherial neural function disable and a supplementary control mean for normal person. It includes brain science, cognitive neuroscience, signal processing, pattern recognition and other interdisciplinary subjects. It has potential application value in automatic control, military and medical fields. In order to obtain efficient signal, we need to deal with the EEG signals. And also machine learning algorithm is a key technology of the brain-computer interface.This work first introduced the brain-computer interface's background, significance, basic principles, system components, classification algorithm and development. Secondly, it analyzed the EEG signals mechanism and principle. Finally, analyzed several types of semi-supervised learning algorithm and applied them into the brain-machine interface.The main content of this paper is the application of the semi-supervised learning algorithm in the brain-machine interface. Then several commonly used semi-supervised learning algorithms were analyzed, and they were introduced into the brain-computer interface system. A semi-supervised framework which based on manifold (LapSVM) was mainly discussed, and the algorithm of Transductive SVM (TSVM), semi-supervised learning by low density separation, and the TSVM using the concave-convex procedure (CCCP-TSVM) was compared. The experimental results showed that the LapSVM was more effective to reduce the complexity of subjects'training than other methods, and it improved brain-computer interface's classification accuracy and laid foundation for the development of brain-computer interface. At the same time, the semi-supervised learning based on manifold can be extended a machine learning algorithm framework, which includes several major machine learning: fully supervised learning, semi-supervised learning, and unsupervised learning. The article used the framework to analyze the EEG signals, and the results highlighted the advantages of LapSVM. At last, we reviewed the work and research results, and also prospected future work still required further study.