he emerging discipline of Affective Computing pursues the development of computers that could interact with their users taking their affective states into account. For example, if a computer could detect when its user is experiencing stress, it could change the colors and sounds of its user interface to try to calm him/her down. Similarly, the pace of instruction in a computer-based training system could be adapted according to the stress level sensed in the pupil. The research described in this paper aims at the development of a stress detection approach based on automatic monitoring of physiological signals in the computer user. The paper describes the three main aspects of our work: experiment setup for physiological sensing, signal processing to detect the affective state and affective recognition using a learning system. Four signals: Galvanic Skin Response (GSR), Blood Volume Pulse (BVP), Pupil Diameter (PD) and Skin Temperature (ST) are monitored and analyzed to differentiate affective states in the user, in a non-invasive fashion. Results indicate that the physiological signals monitored do, in fact, have a strong correlation with the changes in emotional state of our experimental subjects when stress stimuli are applied to the interaction environment.