Decoding the Bimanual Movement from Non-Human Primate Epidural Electrocorticography based on Brain States

Abstract

Recently, advancement of brain signal processing and neuro-technology have enabled us to interact with the brain. With the aid of these techniques, numerous researchers are able to understand and decode the brain signals which have low electrical power. These signals are used for brain-computer interface (BCI) and develop communication systems in which users explicitly manipulate their thought process, instead of motor movements, to control the computers or machines for communications. Thus, the need for these systems is extremely high, mainly to those who faced with physical disabilities. Especially for patients, who left the cognitive functions intact but suffer from paralysis, neuromuscular injuries, or gradual neurodegenerative diseases, which are lack of the voluntary muscular activity, the arm movement tools used in motor BCI have been studied a lot. In arm movement BCI, unimanual studies have been well studied, while bimanual studies have not been well researched because they are relatively complex to interpret brain signals. The brain state of the bimanual movement has been known to be different from that of the unimanual movement, therefore it could be unreasonable to apply the method, which is used in unimanual movement BCI, to bimanual movement BCI. Thus, the needs for the new method has been raised. The purpose of this dissertation is to decode non-human primates' intention and to predict both arms' movement trajectories more precisely. So, I suggest that arm movement BCI should be considered of the different brain states depending on the movement type. In this research, I verified a two-staged method, which applies an effector-sensitive decoder after classified arm movement type, from two rhesus monkey. Using machine learning algorithm, the movement type has been classified and the results of the prediction accuracy have been increased. In following chapters, general concepts of brain signal recording technology, bimanual movement BCI and backgrounds of machine learning are explained in Chapter 1. With this knowledge, in Chapter 2, introduced the new two-staged methods, machine learning methods and experimental methods of the whole research. The results obtained with the new technique are illustrated in Chapter 3, and discussion and future works in Chapter 4.