以腦波為基礎之癱瘓病人溝通輔具技術開發;Development of EEG-Based Communication Assistive Technology for Disable Patients

NCU Institutional Repository > 資訊電機學院 > 電機工程研究所 > 博碩士論文 > Item 987654321/77547

請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/77547

題名:	以腦波為基礎之癱瘓病人溝通輔具技術開發;Development of EEG-Based Communication Assistive Technology for Disable Patients
作者:	許顥騰;Hsu, Hao-Teng
貢獻者:	電機工程學系
關鍵詞:	肌萎縮性脊髓側索硬化症;大腦人機介面;腦電波;穩態視覺誘發電位;前額穩態視覺誘發電位;調適性類神經模糊分類器;Amyotrophic lateral sclerosis (ALS);brain computer interface (BCI);electroencephalography (EEG);steady-state visual evoked potential (SSVEP);frontal SSVEP;adaptive neuron-fuzzy classifier (ANFC)
日期:	2018-04-27
上傳時間:	2018-08-31 14:48:02 (UTC+8)
出版者:	國立中央大學
摘要:	許多受到嚴重運動神經損傷的病患，如肌萎縮性脊髓側索硬化症(amyotrophic lateral sclerosis, ALS),、肌肉萎縮症(muscular dystrophies)、頭部創傷(head trauma)、多發性硬化症(multiple sclerosis)，均難以按照自身意願與外界自由構通。這類病患急需要輔助器來維持他們的日常生活。在最近幾年中，許多研究團隊使用穩態視覺誘發電位(steady-state visual evoked potential, SSVEP)來發展大腦人機介面(brain computer interface, BCI)。穩態視覺誘發電位具有高時間解析度、低量測成本、高資料傳輸率(information transfer rate, ITR)以及可廣泛應用在每個人身上等諸多優點。在傳統上，穩態視覺誘發電位是透過量測使用者的枕葉部分所取得。然而，對於許多人來說，枕葉部位時常有毛髮覆蓋，導致在使用上有較不方便的電極設置以及有較長的穿戴時間。這些使用上的不便利，都會導致穩態視覺型之大腦人機介面系統難以普及應用，更難以適合運動腦區損傷患者使用。並且，穩態視覺誘發電位的相位特徵會有擾動特性，這使得在相位編碼(phase-tagged)穩態視覺型大腦人機介面在注視目標的分類上會有不確定性，導致錯誤率上升。因此，找出可便利設置的穩態視覺誘發電位量測方式與使用可適應性的分類器於相位判定是在穩態視覺誘發型之大腦人機介面設計中相當重要的議題。在本研究中，我們使用前額穩態視覺誘發電位來作為大腦人機介面的控制訊號，以達到輕鬆穿戴之目的。而因著枕葉穩態視覺誘發電位有著振幅-頻率響應特性(amplitude frequency characteristics)，故前額穩態視覺誘發電位在使用上也需要考慮其振幅-頻率響應特性。為了能夠成功實現前額穩態視覺誘發之大腦人機介面，我們研究了年輕人、老年人以及肌萎縮性脊髓側索硬化症病患者的前額穩態視覺誘發電位之振幅-頻率響應特性，並且評估將其訊號作為大腦人機介面之控制訊號的可能性。在這三組受測者中，其平均準確率分別可達96.1 3.2、 91.8 4.2以及81.2 3.7 %。這研究成果顯示，前額穩態視覺誘發電位可成功作為大腦人機介面的控制訊號。為了進一步提升穩態視覺誘發電位之相位辨識，我們使用調適性類神經模糊分類器(adaptive neuron-fuzzy classifier, ANFC)來提高振幅與相位特徵的辨識準確度。在離線分析(off-line analysis)上，其傳統方法與調適性類神經模糊分類器分法的準確度分別為77.26 2.36 % 和95.11 0.90 %。而在即時大腦人機介面的操作上，其平均準確率可達91.7 4.9 %。其研究成果可看到調適性類神經模糊分類可有效提高相位特徵的準確度，並且可應用於上即時大腦人機介面之操作上。論文的研究成果可將穩態視覺誘發電位之大腦人機介面做更有效的落實與推廣，並且成為癱瘓病患有效的構通輔具。 ;Many motor impairment patients, such as amyotrophic lateral sclerosis (ALS), muscular dystrophies, head trauma, and multiple sclerosis, are incapable of communicating with external environments through their free wills. They are in an urgent need of an assistor to maintain their daily activities. In recent years, many research groups have utilized steady-state visual evoked potential (SSVEP), owing to its advantages of high temporal resolution, low cost, high information transfer rate (ITR) and wide suitability to users, to develop brain computer interface (BCI). Conventionally, in order to obtain SSVEP, the EEG signals from parietal-occipital area should be recorded. However, for most people, parietal-occipital area is usually covered with hair which lead to the unpleasant electrode gel setup and increase EEG preparation time. The inconvenient setups of SSVEP-based BCI system were difficulty to promote popularly and were not suitable for motor impairment patients. Moreover, the phase feature of SSVEPs has the characteristics of variations which lead to ambiguity in classifying different gaze targets in the phase-tagged SSVEP-based BCI. Therefore, an easy-preparation for SSVEP recording and a robust and adaptive classifier for phase identification are two crucial issues to develop a practical BCI system. In this dissertation, we utilized frontal SSVEP to implement BCI for achieving the easy-preparation purpose. Owing to the amplitude-frequency characteristics of occipital SSVEP, the amplitude-frequency characteristics of frontal SSVEP should be taken into account. In order to achieve a frontal frequency-coded SSVEP-based BCI, we studied the amplitude-frequency characteristic of frontal SSVEP in young, elderly, and ALS groups, and evaluated its possibility as control signals for BCI applications. The averaged accuracies in operating frontal SSVEP-based BCI in young, elderly, and ALS groups were 96.1 3.2, 91.8 4.2, and 81.2 3.7 %, respectively. The result of this dissertation has achieved the first frontal SSVEP-based BCI in the world. The frontal frequency-coded SSVEP could be an alternative choice to design SSVEP-based BCI. Regarding the robust and adaptive classifier for phase identification, we adopted adaptive neuron-fuzzy classifier (ANFC) to improve the identification of amplitude and phase features in gaze targets and non-gaze conditions. The averaged accuracy in traditional method (preset margin) and ANFC were 77.26 2.36 % and 95.11 0.90 %, respectively, in off-line analysis. In on-line BCI operating, the averaged accuracy in using ANFC was 91.7 4.9 %. The experimental results show that ANFC has ability to improve the accuracy of phase identification, and ANFC is suitable for implementing real-time BCI operating. The achievement provides an effective communication assistor for disable patients.
顯示於類別:	[電機工程研究所] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	204	檢視/開啟

在NCUIR中所有的資料項目都受到原著作權保護.

社群 sharing

資料載入中.....