Learning systems provide adaptation to the learners¢ preferences without taking under consideration the learners¢ current status. The more a learning system exchanges relevant fragments of information about the learner¢s affective status the more it adapts to it. Following this direction, we propose an integrated learning system taking under consideration learners¢ emotional state in order to provide a personalized e-learning system. An extended version of the IEEE Reference Model (WG) LTSA (IEEE 2003) is used for this purpose. The proposed approach is based on the automatic analysis of the learners¢ emotional state providing different learners¢ profiles which are built and maintained by “observing” each learner behavior. As a learner is strongly positively affected to the learning procedure in the presence of an agent, the proposed system has adopted an expressive ECA (Embodied Conversational Agent) which is adapted to the learner¢s emotional states in the duration of the learning procedure.