The United States military is increasingly using infrared (IR) sensors to discriminate the actions and intentions of people being observed in an area of interest; however, there is currently inadequate modeling capability to a priori determine the effectiveness of IR sensors for these types of tasks. The U.S. Army’s NVTherm model is a military and industry standard sensor performance model that estimates target acquisition performance based on both the sensor design parameters and an empirically measured calibration factor to represent human observer performance. Historically the model has been calibrated by presenting static imagery to observers and measuring average probabilities of accomplishing the given task. The task of human activity discrimination, however, presents new challenges to the model, as “activity” inherently implies a dynamic scene where motion cues are essential in accomplishing the task. A series of studies has been completed in order to calibrate NVTherm for the task of human activity discrimination. The challenges involved in representing the human activity task, the establishment of new processing methods, and new standards for defining simple target metrics for complex imagery are discussed. The experiments that have supported the calculation of new calibration parameters are also described. These efforts have brought the U.S. Army significantly closer to having a sensor performance model validated for discriminating human activity with IR sensors.