论文部分内容阅读
人与机器人交互是机器人技术领域、尤其是生活辅助机器人领域的重要课题。本文以辅助老年人、病人和残疾人为应用背景,提出了“智能辅助生活系统”(SAILSystem),并解决了该系统中人的手势识别和日常动作识别两个重要问题。对于手势识别问题,本文采用一个惯性传感器来采集被试验人手指部位活动的信号,运用人工神经网络进行手势捕捉,并应用一个分层隐马尔可夫模型结合前后手势的关联信息,来提高手势识别的准确率。对于动作识别问题,数据来源于位于被试验人一侧的脚面和腰部的两个惯性传感器,并采用多传感器融合方法识别各种日常动作。在对两个传感器的数据进行融合的粗分类之后,细分类应用了隐马尔可夫模型和启发式方法来进一步识别各个动作类型。该穿戴式传感器系统经过实验测试,结果证明了本识别算法的有效性和精确性。
Human-robot interaction is an important issue in the field of robotics, especially in the field of life-supporting robots. Based on the background of assisting the elderly, patients and the disabled, this paper proposes “SAILSystem” and solves two important problems of human gesture recognition and daily motion recognition in this system. For the gesture recognition problem, this paper uses an inertial sensor to collect the signals of the finger of the tested person, and uses artificial neural network to capture the gestures. A hierarchical Hidden Markov Model (HMM) is combined with the related information of the fore-and-aft gestures to improve the gesture recognition The accuracy rate. For motion recognition problems, the data were derived from two inertial sensors located on the instep and waist of the subject and using multisensor fusion to identify various daily actions. After a rough classification of the fusion of the two sensors’ data, the subdivision employs a Hidden Markov Model and a heuristic method to further identify each type of action. The wearable sensor system is experimentally tested, and the result proves the effectiveness and accuracy of this recognition algorithm.