This paper constructs a model of cultural identity by combining the analysis of multimodal discourse in the media fusion perspective and explores how to better construct cultural identity. Firstly, we construct a multimodal cultural identity model in the media convergence perspective from three modalities: text, image, and video combined with cultural identity, so that we can analyze people’s cultural identity through emotion. Secondly, multimodal cross-sectional comparison experiments and dissimilarity experiments are conducted based on two widely used public datasets in the field, CMU-MOSI and CMU-MOSEI. The best experimental results were obtained in the baseline model, where Acc-2 reached 84.8%, F1 value reached 84.5, and MAE of the regression task dropped to 0.548. In the robustness study, when the missing rate reached 44%, Acc-2 was only 65.2%, and the gap between the results and the dichotomous classification accuracy of 76.5% achieved by the model in this paper reached 11.3 percentage points. Finally, through affective analysis, cultural identity is the affirmative embodiment of what is most meaningful to the nation formed by people living together in a national community for a long time, and its core is the identification with the basic values of a nation, which is the spiritual bond that unites this national community.