Word frequency effects in sound change as a consequence of perceptual asymmetries: An exemplar-based model