The invention relates to a micro-expression recognition method based on macro information knowledge migration. The method comprises the following steps: (1) performing kernel function mapping on a macro information sample and a micro expression sample; (2) carrying out local optimization on the macro information sample and the micro expression sample after kernel function mapping, so as to enablethe distance between k1 intra-class nearest neighbor samples of a certain sample to be minimum, and enabling the distance between k2 inter-class nearest neighbor samples of the sample to be maximum; (3) through accumulation of overall alignment local optimization results, enabling the intra-class distance of the whole sample to be minimum and the inter-class distance of the whole sample to be maximum, obtaining a group of projection matrixes, and projecting the macro information sample and the micro expression sample to a common subspace; and (4) carrying out classification through a migrationsupport vector machine model. Compared with an existing micro-expression recognition method, the method provided by the invention has the advantages that the micro-expression recognition performanceis improved, and the robustness is relatively high.