Research on Behavior Score for Patient Support Robot

本研究的目的是建立一种新的行为表征方法。该方法被称为“行为分数”，其灵感来自于代表时变信息的音乐旋律。我们以患者支持机器人为例，对行为评分的有效结构和操作进行了研究。(1)行为分数原型的实现：考虑、定义和实现分数的数据结构、层次结构及其操作集。(2)基于行为评分的多跟踪器选择：为了监测和支持患者，需要对各种多传感器输出进行适当处理，并根据情况检测和跟踪某些特定的特征信号。这种传感器信号的选择和跟踪是基于行为评分实现的。(3)基于行为分数的视觉加工：参考和利用行为分数使视觉加工更加稳健。(4)基于行为评分的患者支持：参考并利用累积的行为评分生成支持患者的计划。(5)基于分数的机器人表达行为：用分数来描述机器人的表达行为是有效的。

The aim of thie research is to establish a novel method to represent behavior. The method is calls as "behavior score"inspired by musical sore that represents time varying information. We took a patient support robot as an example that utilizes the behavior score and investigated effective structure and operations of the score.(1)Realization of behavior score prototype : Data structure and hierarchy of the score and set of operations for them are considered, defined and implemented.(2)Multiple tracker selection based on the behavior score : To monitor and support a patient, various kinds of multiple sensor outputs are properly processed and some specific feature signal should be detected and tracked depending upon situation. Such sensor signal selection and tacking are realized based on the behavior score.(3)Visual processing based on the behavior score : Referencing and utilizing the behavior score make the visual processing more robust.(4)Patient support based on the behavior score : Referencing and utilizing the accumulated behavior score generate a plan to support the patient.(5)Expressive robot behavior based on the score : Applying the score to describe robot's expressive behavior turns out to be effective.