Prediction of Movements by Online Analysis of Electroencephalogram with Dataflow Accelerators
In International Congress on Neurotechnology, Electronics and Informatics, (NEUROTECHNIX), 25.10.-26.10.2014, Rome, SpringerSCITEPRESS – Science and Technology Publications, Oct/2014.
Brain Computer Interfaces (BCIs) allow to use psychophysiological data for a large range of innovative applications. One interesting application for rehabilitation robotics is to modulate exoskeleton controls by predicting movements of a human user before they are actually performed. However, usually BCIs are used mainly in artificial and stationary experimental setups. Reasons for this are, among others, the immobility of the utilized hardware for data acquisition, but also the size of the computing devices that are required for the analysis of
the human electroencephalogram. Therefore, mobile processing devices need to be developed. A problem is often the limited processing power of these devices, especially if there are firm time constraints as in the case of moveme t prediction. Field programmable gate array (FPGA)-based application-specific dataflow accelerators are a possible solution here. In this paper we present the first FPGA-based processing system that is able to predict upcoming movements by analyzing the human electroencephalogram. We evaluate the system regarding computation time and classification performance and show that it can compete with a standard desktop computer.
Brain Computer Interfaces, Mobile Systems, FPGA, Hardware Accelerator