A FRACTAL-BASED ALGORITHM OF EMOTION RECOGNITION FROM EEG USING AROUSAL-VALENCE MODEL 文獻心得

情緒辨識可以從腦波圖有很多地方可以應用，從腦波圖可以讓知道生理內部的情感，在這一篇文獻中，作者提出 fractal dimension（ＦＤ）的運算模型以 Arousal-Valence emotion model來建立2D emotion model，並利用支持向量機(SVM)來分類Arousal、Valence。
使用方法是利用音樂和聲音的刺激來引起受測者的情感，利用IADS 資料庫來實驗 。




The algorithms consist from two parts: feature extraction and classification.  (Lin et al., 2009)
有學者利用 特徵與SVM去分類情緒，
區分出四種情緒:悲傷、憤怒、快樂、喜悅， by (Bos., 2006) using a Binary Linear Fisher’s Discriminant
Analysis, and the emotion states such aspositive/arousal, positive/calm, negative/calm and negative/arousal。

 By applying lifting based wavelet transforms to extract features and Fuzzy C-Means clustering to do classification, sadness, happiness, disgust, and fear emotions were recognized by (Murugappan  et al., 2008).

目前情感識別出的情緒有:悲傷、快樂、難過、厭惡、恐懼、憤怒、喜悅。

此篇PAPER 利用(FD)認識到 negative high aroused, positive high aroused, negative low aroused, and positive low aroused ，以EMOTIV的儀器利用頭腦上的三個點位。

總共點位有128個，在實驗上利用14個，以雜訊干擾最少有三個AF3、F4、FC6，收到訊號後再以SVM分類歸類情緒。

由文獻上知道腦波訊號是非線性和沒有順序性，(FD)模型可以應用於腦波數據上分析(Pradhan and Narayana Dutt, 1993, Lutzenberger et al., 1992, Kulish et al., 2006a, Kulish et al., 2006b)。

可區分積極和消極情緒:

positive and negative emotions, dimensional complexity could be used (Aftanas et al., 1998).

The concentration level of the subjects can be detected by the value of fractal dimension (Wang et al., 2010).

用音樂刺激引誘出情緒的實驗 FD方法:

 Experiments on emotion induction by music stimuli were proposed, and the EEG data were analyzed with fractal dimension based approach in (Sourina et al., 2009a; Sourina et al., 2009b).

作者所提出的方法:

we applied two fractal dimension algorithms for feature extraction, namely Higuchi (Higuchi, 1988) and Box-counting (Falconer, 2003) algorithms as follows. 

CONCLUSION

Higuchi and box-counting algorithms with sliding window were implemented for FD values calculations, and the accuracy of the arousal an valence levels recognition was compared for both algorithms. 

¨negative high aroused (fear)

¨positive high aroused (happy)

¨negative low aroused (sad)

¨ positive low aroused (pleasant)