腦電波情緒識別系統之開發研究

透過腦電波生理訊號理解人們的情緒反應，近年來感性人因工程及神經行銷學，是逐漸重視的研究議題，然而腦電波對情緒分類的影響，及以腦電波為基礎的情緒識別系統的研究尚為欠缺。因此，本研究目的在於：1.評估腦電波訊號對情緒分類的影響；2.開發一以腦電波為基礎的情緒識別系統。本研究使用NeuroSky單通道腦波儀，並將腦電波訊號資料以J48決策樹演算法進行分類，研究結果顯示對情緒分類有影響的腦電波訊號:Attention、Meditation、Theta、Low alpha、High alpha、Low beta、High beta、Low gamma、Mid gamma，而腦電波訊號Delta與情緒影響度較低。此外，透過決策樹分類情緒的組織結構建置的腦電波情緒識別系統，可分類出四類情緒：悲傷、愉悅、厭惡、額外情緒，此系統經實驗驗證準確率為63%。

In recent years, the research of Kansei Engineering and Neurological marketing are getting important and valued a lot. However, it is still an open question how the brainwave influent emotion identification and how to identify emotion from brainwave response. Thus, the aims of the thesis are: 1.) to explore the effect of people’s brainwave on emotion classification; 2.) to develop the emotion identification system based on brainwave. The single-channel electroencephalograph developed by NeuroSky Co. is adopted to record participants’ brainwaves. The brainwave singles are computed with J48 decision tree algorithm to result the emotion classification. Then, the emotion classification results are applied to develop an emotion identification system which is evaluated afterward. The conclusions yield that the brainwaves with significant effects on emotion classification are Attention, Meditation, Theta, Low alpha, High alpha, low beta, High beta, Low gamma and Mid gamma, while the Delta has less significant effects on emotion classification. Moreover, the emotional identification system based on brainwave could be applied to classify four one’s emotions: sadness, pleasure, disgust, rejection emotion. The evaluated accuracy reveals 63%.