The explosive growth of recommender systems has resulted in realization of individualized service as commercial patterns and research prototypes. However, the traditional recommendation approaches are overemphasized the similarity between user preference and items feature. They are completely ignored affectivity that was a crucial factor. Our study focuses on exploring a new affective recommendation approach of semantic associated extension by integrating the Spreading Activation model with knowledge of cognitive psychology for the real-time preference-aware. This paper presents an affectivity-based recommendation approach to eliciting a characteristic sequence consisted of color nodes mapping the relationships between user preference with his mood and items feature. Predominance of our proposal was illustrated through an instantiation of movie recommender system that was developed based on the proposed approach. The testing results of performance show that our affectivity-based recommendation approach outperformed the traditional collaborative filtering approach in terms of the accuracy. This paper also presents a novel insight into exploitation of rich repository of domain-specific knowledge to provide real-time recommendation for user.