Emotion Recognition from Brain Signals Using Hybrid Adaptive Filtering and Higher Order Crossings Analysis 文獻

出處：IEEE TRANSACTIONS ON AFFECTIVE COMPUTING, VOL. 1, NO. 2, JULY-DECEMBER 2010





背景：情緒識別是重要的問題，情緒運算(Affective Computing)提出努力將電腦和一般機器與人互動之能力，表達線索，推斷和證明Emotional Intelligence-相關態度。情緒識別之成功，使機器識別使用者之情緒狀態，並收集情緒之資料進行處理，透過人機界面反應情緒。實現有效的情緒識別，種類繁多的方法和裝置已實施，主要是關於ER臉[4]，[5]，[6]，講話[7]，[8]，信號的自主神經系統（ANS ），即心臟速率和皮膚電反應（GSR）[9]，[10]，[11]。

目的：

1.      developing the novel Hybrid Adaptive Filtering (HAF) for efficiently isolating the emotion-related EEG-characteristics, by applying Genetic Algorithms [13] to the representation of the EEG signal on the Empirical Mode Decomposition [14] domain;

2.      further extending the effectiveness of Higher Order Crossings (HOCs)-based feature vector, initially presented in [15], by structuring it via the oscillatory pattern of the HAF-outputted EEG signal, instead of the EEG signal itself (as used in [15]);

3.      examining the potential of the intrinsic modes of the EEG signal to discriminate between different emotions;

4.      analyzing the classification power of individual EEG channels or combinations of them using the proposed feature extraction method.

0114心得：

作者透過情緒等相關文獻，整理出過去研究情緒的實驗，透過2D Valence-Arousal Model理論，在研究者透過實驗情緒識別(ER)，使機器可以與人互動，表達線索，推斷和證明Emotional Intelligence-相關態度。

作者提出新的特徵擷取方法，為「HAF-HOC」(未看完)為目的項目一，利用臉部圖像以六種基本情緒為樣本，受測16人為慣用右手，9男7女，19~32歲之間。(待續)