基于ADCP-TOP的微表情识别方法

微表情识别的难点是情绪持续时间短,面部动作变化微小。为此,提出一种基于相邻双交叉局部二值模式(ADCP-TOP)的微表情识别方法,针对微表情特点将邻域像素之间关系引入,使得对细节信息提取更丰富,捕捉微小变化能力增强。通过对奇偶位置的采样点分开编码,将结构信息量化,在保证信息量增加的同时使维度减小,并增强鲁棒性。此外,通过面部动作单元划分细粒度感兴趣区域(FROI)提取ADCP-TOP特征,以进一步增强对细节信息提取能力。最后,在SMIC和CASME2数据库中的实验表明,提出的识别方法取得更高识别率。

Micro-expression Recognition Method Based on Adjacent Double Crossover Local Binary Pattern from Three Orthogonal Planes

The difficulty of micro-expression recognition is that the emotional duration is short and the facial movement changes are small. Therefore,a micro-expression recognition method based on the adjacent double crossover local binary pattern from three orthogonal planes(ADCP-TOP)is proposed. According to the characteristics of micro-expressions,the relationship between neighboring pixels is introduced to make the extraction of detailed information richer,and the ability to capture small changes is enhanced. By separately encoding the sampling points of the parity position,the structural information is quantized,the dimension is reduced while the amount of information is increased,and the robustness is enhanced. Furthermore,the ability to extract detail information is further enhanced by extracting ADCP-TOP features from fine-grained regions of interest(FROI)divided by facial action units. Finally experiments on the SMIC and CASME2 micro-expression databases show that the proposed method achieves a higher recognition rate than the existing methods.