【摘要】 针对传统的稀疏编码模型中假定编码残差服从高斯分布,而在实际应用中,编码残差是不可能完全服从固定分布的问题,采用了一种鲁棒稀疏编码求解方法,并将其应用到表情识别中,通过在Cohn-Kana-da数据库上的测试,证明这一基于鲁棒稀疏编码的表情识别方法比传统基于稀疏表达的识别方法(SRC)识别率更高,并且对遮挡和噪声更鲁棒。更多还原

【Abstract】 In traditional sparse coding model,coding residual is assumed to follow Gaussian or Laplacian distribution,but in the actual application the coding residual may not follow an uniform distribution.For this reason,a robust sparse coding calculating method was applied to the facial expression recognition.The experiment on the Cohn-Kanada database shows the facial expression recognition method based on robust sparse coding is more accurate than the one based on traditional SRC,and more robust to corruption and occlusion.更多还原