| DC 欄位 |
值 |
語言 |
| DC.contributor | 電機工程學系 | zh_TW |
| DC.creator | 蕭國彬 | zh_TW |
| DC.creator | Guo-Bin Xioa | en_US |
| dc.date.accessioned | 2019-8-22T07:39:07Z | |
| dc.date.available | 2019-8-22T07:39:07Z | |
| dc.date.issued | 2019 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=106521091 | |
| dc.contributor.department | 電機工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 隨著全球腦科學研究的發展，大腦在應用技術上推陳出新，腦電波的發展提供了有效的工具用於揭開人腦的神祕面紗。「情緒」從古至今都是左右決策、影響勝敗的關鍵。本研究以國科會委任台灣大學執行的情緒計劃所提供的腦波誘發資料，和實驗室自製的8通道腦波機及乾式電極當作實驗設備，蒐集所需要的情緒腦波的訓練資料，從腦波中找出情緒在自主程度暨正負向性腦波的特殊特徵驗證實驗確效性，由於不同年齡、不同性別、不同慣用手上，大腦在不同區塊活躍度有所不同，藉由深度學習來訓練一組適合各年齡層，不同性別都適用的人工智慧模神經網路模型，以學校師生男女共15人的腦波當成訓練資料，給予每位受試者觀看20分鐘的腦波刺激材料，同時利用專門為深度學習打造，具備有GPU模組的TX2進行有效率的深度學習，以雙線程為主軸，當在接收腦波的同時，即時將腦波經處理置入訓練好的神經網路模型，快速辨識受測者當下的情緒在自主程度、正負向性的情緒指標。在準確率方面，正負向性準確率經交叉驗證後的準確度達到81.48%，自主程度準確度達到74.18%，透過此8個通道腦波規格及創新的資料預處理方法來達成過去必須要32通道或是64通道所訓練出來的準確率及成果，完成在應用上更具輕便性以及輕巧性的系統，致力於減少通道數及系統的精簡性。在應用層面，為實體化辨識的情緒指標，引進虛擬實境做為媒介跟時代接軌，透過立體環境模擬遠端視訊，同時讓另一端知道自己真實的情緒狀態，將冷冰冰的理論實現到現代應用設備上。 | zh_TW |
| dc.description.abstract | With the developments of research in brain science and the novel technologies of brain-related applications technology, electroencephalography (EEG) provides an effective tool to unveil the mystery of human brain. Especially, ＂emotion＂ is the most attractive topic which can influence decision and be the key to victory or defeat. In this dissertation, emotion stimulation materials are provided by the emotional plan of National Science Council, which is implemented by the National Taiwan University. Eight-channel EEG machine and high-sensitivity dry electrodes are used as the experimental equipment to collect the emotional EEG training data. Human emotions were represented in valence and dominance domains. The relations between EEG temporal-frequency features and emotion markers verified the the correctness of the experiment. Since brain activities are different across individuals owing to their ages and habitual hand, neural network is a suitable choice for our emotion classification study. In the study, fifteen subjects were recruited to view a twenty-minute emotion stimulation material. EEG data measured from the fifteen subjects were used as training data for deep-learning neural network. After network training, real-time recognition of subjects’ emotions were performed on TX2 platform whose GPU module was configured by the pre-trained Deep Learning neural network. In our system, simultaneous EEG data recording and real-time emotion recognition was available with the use of double-threading system. The accuracies of valence and dominance indexes have achieved 81.48% and 74.18%, respectively. The research results of our eight-channel EEG study in this thesis have attained comparable results compared to other previous literatures which utilized 32- or 64-channel EEG system in emotion studies. In addition, we also introduced virtual reality (VR) to integrate with our proposed eight-channel EEG emotion detection system. The VR-EEG emotion detection system has already implemented in 3D environment and combined with social communication software which can achieve a variety of entertainment applications in future studies. | en_US |
| DC.subject | 深度學習 | zh_TW |
| DC.subject | 腦波 | zh_TW |
| DC.subject | 情緒辨識 | zh_TW |
| DC.subject | 人工智慧 | zh_TW |
| DC.subject | 虛擬實境 | zh_TW |
| DC.subject | Deep Learning | en_US |
| DC.subject | Electroencephalography(EEG) | en_US |
| DC.subject | emotional recognition | en_US |
| DC.subject | artificial intelligence | en_US |
| DC.subject | virtual reality | en_US |
| DC.title | 深度學習於腦波情緒辨識研究 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |