P300視覺事件誘發電位（Event-Related Potential, ERP）是腦機介面（Brain-Computer Interface, BCI）經常使用之實現方法；由於P300不需要大量資料之訓練與建模，所以使用者無需長時間之訓練就能熟悉操作系統；因此， P300腦機介面技術已廣泛地應用於協助脊椎損傷或中風之使用者使用電子裝置。然而腦波信號之變化量較小，往往必須將信號做5~10次之疊加才能進行判讀，造成P300之腦機介面系統需要較長之判讀時間；這對於資料傳輸速度需求較低之拼字系統應用上是可行的，然而當應用於動態機器人控制上，判讀即時性是具有挑戰性的。有鑑於此，本論文提出一以時間位移相關性（Time-shift correlation）為基礎之P300判讀演算法，此演算法可解決事件誘發電位各波峰每次出現時間偏移以及振幅大小不一之問題。同時，本文以時間位移關聯因子序列為輸入，並以類神經網路進行P300信號判讀，以達到快速判讀P300誘發電位。在實作上，本研究以自製之三通道（Cz、Pz、Oz）腦波量測儀量取腦波訊號，並將上述之時間位移相關性演算法實現於微處理器，最後應用於大型人形機器人之步行控制。
本研究所提出之操作模式包含快速模式與精確模式；當使用者操作機器人於寬敞之空間，以快速模式以較快的判斷速度對機器人下達步行指令；當機器人處於較狹小之空間，則以準確度較高之精確模式下達控制指令，協助機器人於窄巷之脫困並避免過多誤判而產生碰撞。實驗結果顯示快速模式在100次實驗中之成功率達87%，平均每分鐘可產生30.3 次正確指令；其中有87次成功判讀，23次誤判，0次無法判斷。然而在精確模式之下，成功率提升為95%，誤判率為5%，但因辨識法則條件較為嚴苛，使得平均每分鐘可產生15.2 次正確指令。

P300 is an event-related potential (ERP), and it is a popular brain-computer interface for detecting visual evocation stimulus. P300-based system has the advantages of using small amount of user’s data for training and modeling. Hence, P300 is feasible for practical applications without requiring long term training. Therefore, P300 techniques have been widely applied to control electronic devices for spinal cord injury and stroke patients.
However, the electroencephalography (EEG) signal level is very small, and usually the P300 can be observed from 5 to 10 times accumulation of EEG signals. As a consequence, P300 needs a longer time for recognition. Such a phenomenon could be available for low information transfer rate systems, such as computer spelling interfaces. Nevertheless, it is quite challenge to use P300 to control a dynamic movement system, such as robots.
To improve the information transfer rate, this work proposes a time-shift correlation approach for recognizing P300 systems. This approach is capable of dealing with the wave form variations from P300 peak time and voltage ranges. Moreover, the time-shift correlation series data is collected as the input nodes of the neural network (NN), and the classification of four LED visual stimuli is selected as the output node of the NN. Practically, a three-channel EEG instrument for collecting the EEG signals from Cz, Pz and Oz electrodes. The three-channel EEG signals are further processed to extract the time-shift correlation series data. The classification is used to control an adult-size humanoid robot.
Two modes including fast-recognition mode (FM) and accuracy-recognition mode (AM) are proposed for different operation conditions. When a robot walks in a spacious area, the FM is used to control the robot with a higher information transfer rate. When a robot walks in a crowded area, the AM is used for the consideration of high accuracy of recognition to reduce the opportunity of collision. The experimental results showed that in 100 trials, the accuracy rate of FM is 87%; the average information transfer rate is 30.3 commands/ min., and there are 13 misjudgments. In addition, the accuracy rate is improved to 95% for the AM, and the average information transfer rate decreases to 15.2 commands/ min. due to strict recognition constrains.

參考文獻
[1]H. Cecotti and A. Graser, “Convolutional neural networks for P300 detection with application to brain-computer interfaces,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 33, Issue: 3, pp. 433 - 445, 2011.
[2]I. Iturrate, J. M. Antelis, A. Kubler, and J. Minguez, “A noninvasive brain-actuated wheelchair based on a P300 neurophysiological protocol and automated navigation,” IEEE Transactions on Robotics, Vol. 25 , Issue: 3, pp. 614 - 627, 2009.
[3]D. J. Krusienski, E. W. Sellers, F. Cabestaing, S. Bayoudh, D. J. McFarland, and T. M. Vaughan. “A comparison of classification techniques for the P300 speller,” Journal of Neural Engineering, vol. 3, pp. 299–305, 2006.
[4]J. Long, Y. Li, H. Wang, T. Yu, J. Pan, and F. Li, “A hybrid brain computer interface to control the direction and speed of a simulated or real wheelchair,” IEEE Transactions on Neural Systems and Rehabilitation Engineering, Vol. 20, Issue: 5, pp. 720 - 729, 2012.
[5]M. Li, W. Li, J. Zhao, Q. Meng, F. Sun, and G. Chen, “An adaptive P300 model for controlling a humanoid robot with mind,” IEEE International Conference on Robotics and Biomimetics (ROBIO), pp. 1390 - 1395, 2013.
[6]Y. Li, J. Pan, F. Wang, and Z. Yu, “A hybrid BCI system combining P300 and SSVEP and itsapplication to wheelchair control,” IEEE Transactions on Biomedical Engineering, Vol. 60, Issue, 11 , pp.3156 – 3166, 2013.
[7]D. Puanhvuan and Y. Wongsawat, “Illuminant effect on LCD and LED stimulators for P300-based brain-controlled wheelchair,” Biomedical Engineering International Conference (BMEiCON), pp. 254 - 257, 2011.
[8]D. Puanhvuan and Y. Wongsawat, “Semi-automatic P300-based brain-controlled wheelchair,” ICME International Conference on Complex Medical Engineering (CME), pp. 455 - 460, 2012.
[9]I. Pathirage, K. Khokar, E. Klay, R. Alqasemi, and R. Dubey, “A vision based P300 brain computer interface for grasping using a wheelchair-mounted robotic arm,” IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM), pp. 188 - 193, 2013.
[10]B. Rebsamen, E. Burdet , C. Guan , H. Zhang, C. L. Teo, Q. Zeng, M. Ang, and C. Laugier, “A brain-controlled wheelchair based on P300 and path guidance,” BioRob, The First IEEE/RAS-EMBS International Conference on Biomedical Robotics and Biomechatronics, pp. 1101 – 1106, 2006.
[11]A. Turnip, K. S. Hong, and S. Yoon, “Adaptive nonlinear principle component analysis based multilayer neural network for P300 detection,” IEEE/SICE International Symposium on System Integration (SII), pp. 96 – 99, 2011.
[12]B. B. Winter and J. G. Webster, “Driven-right-leg circuit design”, IEEE Transactions of Biomedical Engineering, pp.62-66,1983.
[13]F. Xin'an, L. Bi, Y. Li, K. Jie, and H. Ding, “A P300 brain-computer interface for controlling a mobile robot by issuing a motion command,” ICME International Conference on Complex Medical Engineering (CME), pp. 4707 – 710, 2013.
[14]Y. C. Yu, A. Nawroj, S. Wang, and L. Gabel, “Mobile robot navigation through a brain computer interface,” Signal Processing in Medicine and Biology Symposium (SPMB), 2012 IEEE , pp. 1 – 5, 2012.
[15]S. J. Yun, M. C. Lee, and S. B. Cho, “P300 BCI based planning behavior selection network for humanoid robot control,” Ninth International Conference on Natural Computation (ICNC), pp. 354 – 358, 2013.
[16]J. C. Zhang, Y. Q. Xu, and L. Yao, “P300 detection using boosting neural networks with application to BCI,” IEEE/ICME International Conference on Complex Medical Engineering, pp. 1526 - 1530, 2007.
[17]連育德，「以平滑支撐向量迴歸為基礎之大型雙足機器人類人倒單擺軌跡規劃」，碩士論文，國立台灣科技大學，民國102年。
[18]康志豪，「P300腦機介面拼字器之線上校正系統開發」，碩士論文，中原大學，民國99年。
[19]翁瑞宗，「腦機介面之P300拼字器開發及特徵分析」，碩士論文，中原大學，民國98年。
[20]劉維旻，「無線感測心電圖量測系統設計與應用」，碩士論文，中央大學，民國97年。
[21]鄭偉銘，「可重組構之多重生理訊號量測放大器」，碩士論文，國立台灣科技大學，民國100年。
[22]ALLDATASHEET.COM（2014）。2014年3月1日，取自 http://www.alldatasheet.com/datasheet-pdf/pdf/100457/BURR-BROWN/DCV010505.html
[23]ALLDATASHEET.COM（2014）。2014年3月1日，取自http://www.alldatasheet.com/datasheet-pdf/pdf/48093/AD/AD623.html
[24]維基百科（2014）。2014年6月15日，取自http://en.wikipedia.org/wiki/File:Eeg_alpha.svg
[25]維基百科（2014）。2014年6月15日，取自http://en.wikipedia.org/wiki/File:Eeg_beta.svg