An Exploration Of Brain-computer Control Method Of Multi-joint Robot Arm

With the development of computer technology, electroencephalogram and pattern recognition, there are more studies on brain-computer interface (BCI) technology. An individual can send information to the external world through a BCI, independent of one's normal output ways of nerves and muscles. The disable can use a BCI to control auxiliary equipment such as a wheelchair or an artificial limb. What's more, a BCI can offer auxiliary rest and external strength for a normal people.The present paper mainly studies the application of BCI in the control of a multi-joint robot arm. Firstly, the main work in and abroad is summarized by reading lots of articles and the latest progress in this realm is followed. Secondly, a P300 paradigm and human-machine interface for the control of robot arm is designed. The effects of stimulus time and inter-stimulus time (ISI) on correct rate and time of spelling one character are studied. Furthermore an optimized parameter configuration is selected so that a subject can control the robot arm better. The stimulus time can be reduced to shorten the time for one character when the flash repetition is set to a certain number. The best parameter configuration is different according to different subjects. Through adequate training, a subject can select characters faster while the correct rate is acceptable. A control command can be recognized and executed in 3.3s, and the correct rate is above 80%. Then a motor imagery (MI) paradigm for robot arm control is designed where Movement of 4 joints is controlled by three kinds of motor imagery signals and autoregressive (AR) model is applied to the extraction of frequency feature. Off-line analysis is performed on two-class MI task and three-class MI task where effects of the selection of motor imagery time window to the feature extraction are investigated. Subject-dependent system parameters and classifiers are selected and applied in online experiments. According to the online results problems in the experiment are analyzed and improvements to the online experiment design are proposed. Finally a prospection of future work is made.