In this paper, we address the limitations of existing hypergraph representation methods in the interactive features of nodes within hyperedges, proposing a novel framework, Motif-enhanced hypergraph neural network (MHGNN). MHGNN leverages the flexibility of motifs and their advantage in characterizing complex node dependencies to describe node behaviors within hyperedges, while exploring the interactions among motifs that reflect local substructural features, thereby compensating for the missing local substructures in hyperedges. Specifically, the framework constructs a hypergraph based on motif substructures and assigns topological and semantic features to the motifs. Subsequently, it extracts motif structural features from two perspectives: one is a node-centric view for capturing the behavioral patterns of neighboring nodes and the other is a hyperedge-centric view for capturing the variations in motifs across hyperedges. Finally, a motif interaction fusion mechanism is designed to extract interaction features based on the differences and commonalities of motif types, enabling better learning of the intrinsic properties of motifs and their interaction behaviors. The results of extensive experiments on three benchmark datasets that MHGNN performs better than existing methods. The promising results on node classification demonstrate the effectiveness of enhancing motif interaction information in improving model performance.