情感计算的一个重要任务是情感建模。提出了在人脸情感的视觉识别范畴中基于PAD理论的情感建模。根据Mehrabian提出的PAD3维情感理论,建立了EBM(emotional block model)模型,进行了非典型情感识别的尝试。采用88特征点的Gabor特征和SVM算法在Cohn-Kanade数据集上进行了非典型情感识别以及典型情感识别的实验,并就典型情感的识别与基本情感模型比较。实验结果表明,无论是识别非典型情感还是典型情感,基于PAD理论建立的情感模型都是可靠的。在会聚度高的情感子空间上的识别率比会聚度低的情感子空间高。
An impertant task of affective computing is to build computable emotional models. In our study, PAD theory is used and EBM (emotional block model) is built and verified in facial emotion recognition area. 88 points based Gabor feature and SVM (support. vector machine) classifier are used to verify this model on Cohn-Kanade dataset. Non-basic and basic emotions are recognized with EBM model in our experiment, and the advantage and disadvantage are compared with PAD based models and traditional basic emotional models. Experimental results show that EBM is reliable. The result is better in high-convergent emotional block than in low-convergent emotional block.