The need to provide computers with the ability to estimate the affective state of their users is a major requirement for the practical implementation of Affective Computing concepts. This research aims at sensing and rceognizing typical negative emotional states, especially "stress", when the user is interacting with the computer. An integrated hardware - software setup has been developed to achieve automatic assessment of the affective status of a computer user. A computer-based "Paced Stroop Test" is designed to act as a stimulus to elicit emotional stress in the subject. Four signals: Blood Volume Pulse (BVP), Galvanic Skin Response (GSR), Pupil Diameter (PD) and Skin Temperature (ST) are monitored and analyzed to differentiate affective states in the user. Several signal processing techniques are applied to the signals collected to extract the most relevant features in the physiological responses and feed them into learning systems, to accomplish the affective state classification. Three learning algorithms are applied to this classification process and their performance is compared. Results indicate that the physiological signals monitored do, in fact, have a strong correlation with the changes in emotional state of our experimental subjects.