Michael has developed neural network and computer vision models that explain the cross-domain transfer for the recognition of dynamic facial expressions with different head shapes. In psychophysical experiments, he has investigated the influence of the basic head shape on expressions recognition.