特征提取是面部表情分类识别的研究重点。针对原始局部方向模式(Local Directional Pattern, LDP)特征提取速率缓慢的问题, 对LDP的编码方案进行改进, 设计了nLDP(new Local Directional Pattern)算子。选择Kirsch算子的 4个方向模板来获取边缘响应值, 然后将正的边缘响应置为1, 负的边缘响应置为0, 从而获得nLDP特征表示, 最后采用支持向量机(SVM)对表情进行识别。实验结果验证了提出的nLDP算子在保证表情识别准确率的同时, 有效地提高了表情识别的速率。

Feature extraction is the research emphasis of facial expression recognition. Specific to the low speed of extracting features when using original Local Directional Pattern(LDP), the coding scheme of LDP was imprved, and a new Local Directional Pattern (nLDP) operator was designed. By choosing four direction template of Kirsch operator to obtain the edge response value, and then directly setting the positive edge response value to 1, the negative value to 0, the nLDP features were gained. Finally, support vector machine (SVM) was used to classify facial expressions. The experimental result shows that the proposed nLDP operator can effectively improve the speed of facial expression recognition and ensure the recognition accuracy.