A person's face discloses important information about their affective state.\nAlthough there has been extensive research on recognition of facial\nexpressions, the performance of existing approaches is challenged by facial\nocclusions. Facial occlusions are often treated as noise and discarded in\nrecognition of affective states. However, hand over face occlusions can provide\nadditional information for recognition of some affective states such as\ncuriosity, frustration and boredom. One of the reasons that this problem has\nnot gained attention is the lack of naturalistic occluded faces that contain\nhand over face occlusions as well as other types of occlusions. Traditional\napproaches for obtaining affective data are time demanding and expensive, which\nlimits researchers in affective computing to work on small datasets. This\nlimitation affects the generalizability of models and deprives researchers from\ntaking advantage of recent advances in deep learning that have shown great\nsuccess in many fields but require large volumes of data. In this paper, we\nfirst introduce a novel framework for synthesizing naturalistic facial\nocclusions from an initial dataset of non-occluded faces and separate images of\nhands, reducing the costly process of data collection and annotation. We then\npropose a model for facial occlusion type recognition to differentiate between\nhand over face occlusions and other types of occlusions such as scarves, hair,\nglasses and objects. Finally, we present a model to localize hand over face\nocclusions and identify the occluded regions of the face.\n