Human Computer Interaction (HCI) can be made more efficient if the interactive systems are able to respond to the users' emotional state. The foremost task for designing such systems is to recognize the users' emotional state during interaction. Most of the interactive systems, now a days, are being made touch enabled. In this work, we propose a model to recognize the emotional state of the users of touchscreen devices. We propose to compute the affective state of the users from 2D screen gesture using the number of touch events and pressure generated for each event as the only two features. No extra hardware setup is required for the computation. Machine learning technique was used for the classification. Four discriminative models, namely the Naïve Bayes, K-Nearest Neighbor (KNN), Decision Tree and Support Vector Machine (SVM) were explored, with SVM giving the highest accuracy of 96.75%.