新穎腦波移動想像分類器及其應用於機械手臂控制

Abstract

腦機介面(Brain computer interface, BCI)是大腦與電腦或其他裝置之間一個有效的溝通橋樑。許多生理訊號都可用於操作腦機介面，其中，運動想像(Motor imagery, MI)已被證明是操作腦機介面的一種有效方式。近年來，在很多關於使用運動想像操作腦機介面的研究中，常見的問題有：正確率不高、訓練的時間太長、無法正確辨認使用者的操作指令以及計算時間太長導致系統反應較慢等問題。本論文提出一新穎SUB-LDA分類器系統架構，不僅提高運動想像判斷的正確率及有效辨認使用者的操作指令，而且也能維持系統的計算效率，正確率達到85%左右。此外，本論文更進一步將此開發出的架構嵌入於無線可攜式腦機介面(Mindo)，將此技術進而應用到控制機械手臂，讓雙手不便的人更加方便的可以自己生活。

Brain computer interface (BCI) serves as the bridge to link human brain and a computer for the scientific/engineering applications. Many physiological signals can be used for the operation of brain-machine interface, which Motor Imagery (MI) has been proven to be an effective way to operate brain-computer interface. Recently on using BCI in terms of the MI studies, apparent problems can be included as common problems are: The accuracy of misclassification is too high, the training time is too long, and calculation time is too long to result in a slower system response. This paper presents a novel LDA ensemble classifier system architecture, not only improve the accuracy of judgments and effective movement of imagination to identify the user's instructions, but also to maintain the computational efficiency of the system. The accuracy is about 85%. In addition, the proposed framework, which both possesses high classification accuracy and fast execution time, could be applied to wireless portable brain computer interface – Mindo, and uses this technology to further manipulate robotic arm in order to facilitate the disabled and the elderly for making life more convenient.