To study the influence of subject emotions on risk communication under emergencies, this paper adopts the Rank Dependent Expected Utility (RDEU) theory to define sentiment functions. Using a metro emergency as an example, a game model of risk communication between the public and the management is constructed, then a Nash equilibrium solution analysis is conducted, and finally a numerical simulation using MATLAB is carried out to investigate the optimal strategy for risk communication in different situations. The results show that emotions can affect the risk communication behaviour of the emergency to a certain extent, but are not a determining factor. When both sides of the game have emotions, the pessimistic emotional state will have a greater impact on decision-making compared to the optimistic emotional state. In addition, it is found that opinion leaders significantly influence the choice of risk communication strategies of game players. This study provides theoretical and methodological support for risk communication in emergencies.