Towards Word Semantics from Multi-modal Acoustico-Motor Integration: Application of the Bijama Model to the Setting of Action-Dependant Phonetic Representations