Computational Intelligence for Community-centric Robot Partners

• 作 者: Janos Botzheim
• 机 构: TokyoMetropolitanUniversity,Japan
• 出 处: 第三届先进计算智能与智能信息学国际研讨会(The Third International Workshop on Adv
• 发表日期: 2013年10期

　　We develop robot partners based on smart phones for elderly care. We apply Kinect and iPhone sensors for measuring distance, images, sound and others related to the human beings. For community-centric robot partner, multimodal recognition is an important task. It includes the recognition of human and objects, face recognition, voice recognition, and gesture recognition. Computational intelligence techniques can be used for solving these problems. In the presentation we deal with structured learning composed of unsupervised learning (by growing neural gas or self-organizing maps) and spatiotemporal learning (by spiking neural networks). Another important task is to estimate human behaviors, including human transportation modes, and interaction modes, and to analyze daily life log. We propose several types of structured learning for solving these problems. We show experimental results, and discuss the future direction of research on community-centric robot partners.