Multimodal saliency-based bottom-up attention a framework for the humanoid robot iCub | IEEE Conference Publication | IEEE Xplore