基于选择性集成分类器的面部表情识别研究*

为了提高面部表情的分类识别性能，基于集成学习理论，提出了一种二次优化选择性(Quadratic Optimization Choice, QOC)集成分类模型。首先，对于9个基分类器，依据性能进行排序，选择前30%的基分类器作为集成模型的候选基分类器。其次，依据组合规则产生集成模型簇。最后，对集成模型簇进行二次优化选择，选择具有最小泛化误差的集成分类器的子集，从而确定最优集成分类模型。为了验证QOC集成分类模型的性能，选择采用最大值、最小值和均值规则的集成模型作为对比模型，实验结果表明：相对基分类器，QOC集成分类模型取得了较好的分类效果，尤其是对于识别率较差的悲伤表情类，平均识别率提升了21.11%。相对于非选择性集成模型，QOC集成分类模型识别性能也有显著提高。

Facial Expression Recognition Research Based on Selective Ensemble Classifiers

In order to improve the classification performance of facial expressions, this paper proposed a new Quadratic Optimization Choice(QOC) ensemble classification model based on the theory of ensemble learning. Firstly, for the nine base classifiers, according to the performance of sorting, selected the top 30% classifiers as candidate base classifier of ensemble model. Secondly, according to the rules of combination created ensemble model cluster. Finally, the ensemble model cluster was optimized by two times, and selected the subset of ensemble classifier with minimal generalization error. In order to verify the performance of QOC ensemble classification model, to use maximum, minimum and mean value rule as a model for comparison. The experimental results show that the relative base classifier, QOC classification model has achieved good classification results, especially for the poor sad expression recognition rate, the average recognition rate increased up to 21.11%. Compared with the non selective ensemble model, the recognition performance of the QOC ensemble classification model is also significantly improved.