Video lectures have become the most used method of distance learning. But, differently from modern Intelligent Tutoring Systems, video lectures lack the capability to adapt its contents to the student needs. Basically because video lectures are passive to the student. In this thesis, we present an intelligent recommendation system based on video lectures for distance learning. We address the problem of lack of interactivity between the student and the video lecture by passively monitoring the student with sensors. We developed a novel student model that can output theengagement of the student towards the system using the information provided by these sensors. Our system is capable of adapting the flow of the video lecture based on the reported output of our student model. To be able to monitor the engagement of a student, we developed a sensor based on a infrared webcam that can captureeye features, like pupil diameter and blink rate. Using this sensor and our model, we can achieve an accuracy higher than 80% for the engagement of students, compared to the feedback provided by teachers. We also present a model to measure the engagement of students in a video lecture based on how much time the student spent in the lecture.