Sequential Human Behavior Recognition for Cooking-Support Robots

Recent advances in information technology are making electric household appliances computerized and networked. If our environments could intuit our activities, e.g., by sensors, novel services taking anticipated actions into account would become possible. We propose activity recognition that infers a subject’s next action based on previously observed behaviors. We developed a cooking-support robot that suggests by voice and gesture what the subject may want to do next. Experimental results confirmed feasibility of the inference and the quality of support.