The Study And Implementation Of Process Deployable Distributed Computation Platform For Motor Brain-Computation Interface

Brain-computer interface (BCI) technology provides a direct communication and control channel between the brain and external devices. With the development of human brain science, BCI technology is gaining more and more momentum. Among the procedures of BCI researches, signal analysis occupies the core position. Because of high throughput of neural signals, the calculated amount is huge. And it increased even more as studies getting further and deeper. Whereas thanks to the complexity of BCI research background, a lot of miscommunications appear and lead to problems such as unknown demands and unstable systems, which hinders the development of experimental platform and further studies of BCI. Therefore, it is of vital importance to build a computation platform providing general solutions to commonly problems on BCI computations.In this article, in order to solve above questions, some requirements analyses about BCI’s computation has been concluded in the development, maintenance and upgrades of a non-human primate experimental platform, which leads a general computation platform aimed at on-line decoding has been constructed. This platform decomposes complex computational processes into organizations of simpler ones, and provides services at customizing on-line decoding systems for different calculation needs, with the help of a graphical computational process deploy system.On behalf of computing performance, distributed computation technologies such as Hadoop are introduced in this article to build a platform mainly for off-line computations services of big data processing and high-quality model computation. Based on this platform, distributed algorithms of generalized regression neural network and k-nearest neighbors have been implemented. The performance analysis is provided as well.In the end of this article, the proposed computation systems have been placed in the real scenario of monkey motor decoding experiment for analyzes and discussions of performance on interactivity, scalability and efficiency.