Emotions play a very important role in the learning process. Affective computing studies try to identify users’ affective state, as emotion, using affect models and affect detection techniques, in order to improve human-computer interactions, as in a learning environment. The Internet explosion makes a huge volume of information, including learning objects data, available. In this scenario, recommendation systems help users by selecting and suggesting probable interesting items, dealing with large data availability and decision making problems, and customizing users’ interaction. In u-learning context, students could learn anywhere and anytime, having different options of data objects available. Since different students have different preferences and learning styles, personalization becomes an important feature in u-learning systems. Considering all this, the authors propose the Affective-Recommender, a learning object recommendation system. In this chapter, they describe the system’s requirements and architecture, focusing on affect detection and the recommendation algorithm, an example of use case, and results of system implementation over Moodle LMS.