基于选择性集成的情感识别

Emotion recognition based on selective ensemble

当前,情感识别已经成为情感计算中一个重要研究问题。传统的情感识别方法有人工神经网络(ANN)的情感识别、模糊集的情感识别、支持向量机的情感识别和隐马尔可夫模型(HMM)的情感识别等。将选择性集成的方法应用到情感识别中,该方法通过训练数据集的随机数抽取、训练,得到一批候选分类器,并通过差异性计算,挑选出大于平均差异性水平的分类器用来做最终情感识别。实验表明,该方法比传统的识别方法以及bagging集成方法的效果都好,能有效地提高情感识别的精度。

Emotion recognition is a key problem in affective computing. It is usually studied based on facial and audio information with methodologies, such as artificial neural network （ANN）, fuzzy set, support vector machine （SVM）, hidden Markov model （HMM）, etc. A method of selective ensemble was used in emotion recognition; through the random extraction and training of the training data set, the classifiers whose diversities are over the av- erage level were chosen for recognition using difference calculation. Simulation results show that the method, which effectively promotes the accuracy of emotion recognition, is better than the method of single classifier and even the bagging.