Abstract: This review systematically examines the research progress in digital human facial expression transfer technologies, focusing on three key components: 3D face representation, facial feature extraction, and expression transfer methods. The paper traces technological evolution from explicit 3D morphable models to implicit neural network representations, analyzing critical challenges including expression-identity disentanglement and cross-dimensional expression mapping. Current technical limitations are identified as restricted expressiveness of 3D face models, susceptibility of feature extraction to environmental interferences, and insufficient cross-subject adaptability. Based on comprehensive literature analysis, the review proposes developing deformable face modeling methods with enhanced local expressiveness, exploring precise expression disentanglement techniques based on convolutional implicit neural networks, and creating expression transfer frameworks that integrate implicit neural networks with nonlinear face representations. These innovative directions will help simplify transfer processes, strengthen system generalization capabilities, and ultimately achieve more natural and efficient digital human facial expression generation, advancing digital human technology applications across military training, education, psychological relief, and entertainment industries.