Adaptive Trajectory Generation via Visual and Voice Feedback for Automated Robotic Therapeutic Massage Applications

• Main Authors: Song-Yong Luo, 羅嵩詠
• Other Authors: 羅仁權
• Format: Others
• Language: en_US
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/dm54bx

碩士 === 國立臺灣大學 === 電機工程學研究所 === 107 === Therapeutic massage is an ancient technique, which benefits muscle relaxation, blood circulation, and the relief of fatigue accumulation on the body. After leaving factory, the conventional massage tools have the preprogrammed trajectory in each setting mode, and can''t respond to users'' condition immediately. In contrast to the conventional ones, robotic massage owns the potential to adjust the massage trajectory and applied force at once according to how users react, and performs the massage in a more suitable manner. Currently, most of the trajectory generation of the robotic massages are achieved by teaching by touching or manipulating the human robot interface, meaning that robotic massage is not automated yet. As a solution, machine vision can be introduced, bringing the perception system to the robot. We use the RGB-D camera (Xtion) to capture the depth images of human''s shoulder and back and use these images to generate the 3D model. Besides, providing that people with the similar bode shape, their distribution of acupressure points are alike and there are some correspondences among them. If a database is built with the standard models recording the correct massage trajectory in it, and these standard models belong to people with different body shape. By following these three steps: comparing user''s scanned 3D model with the standard models in the database, finding the most similar one, mapping the massage trajectory onto the user''s model. An approximate massage trajectory is generated. We use the image morthing algorithm in image processing as the way of mapping, which can quickly find the acupressure points distribution of the user. The speech system will also be discussed in subsequent chapters. After listening to and understanding the user''s semantics and needs, the voice system integrates it with the recognized acupressure points to obtain a massage trajectory that meets the user''s needs. In addition, it takes only 6 seconds to generate a massage trajectory from body contour recognition, acupoint distribution positioning, and combined with voice information. Therefore, if the user has the need to adjust the seat and sitting position. The user can pause the massage, adjust the seat and the posture, and then continue to receive massage from the robot, no need to re-correct the camera position, and spend a lot of waiting time.