With the breakthrough advancements in large-scale AI interaction models, the application of emotional interaction in anthropomorphic AI systems across various scenarios such as caregiving and entertainment has garnered increasing attention. This paper, grounded in the principles of interpersonal emotional interaction, analyzes emotional interaction patterns based on human-machine intimate relationships. It proposes an emotional interaction model comprising three modules: perception, behavior, and optimization regulation. The study identifies five key principles of emotional interaction: accuracy, naturalness, personalization, adaptability, and motivation. An initial validation of these principles was conducted using a constructed database. This exploratory study provides valuable insights for the application of emotional interaction in AI systems.