How do users perceive multimodal expressions of affects?