If you have any problems related to the accessibility of any content (or if you want to request that a specific publication be accessible), please contact (firstname.lastname@example.org).
Prosthesis Design and Object Recognition Based Grasping of a 3D Printed Anthropomorphic Artificial Hand
Electrical and Biomedical Engineering
AltmetricsView Usage Statistics
This work presents the design of a cost effective, open-source, 3D printed prosthetic hand with the implementation of object recognition in order to reduce the cognitive burden of amputees through automated grasping. Due to the high cost of modern, active, prostheses, an open-source 3D printable anthropomorphic artificial hand is selected for the purpose of designing a prosthesis.Through the use of image processing, some current prosthetic hands are capable of selecting automated grasps via object recognition. However, many of these methods are computationally heavy, and require large amount of training data for daily usage. Such computationally heavy methods also utilize myoelectric control, which requires training for an extended period of time. In this work, by utilizing intrinsic information of an object, such as color and edge information, the computational time is reduced while still providing the user with automated grasping, without the need for myoelectric control. In future works, illumination invariant techniques may be used to increase the efficiency of color constancy.In addition to this work, modifications and functional enhancements are made to the open-source artificial hand design, called InMoov, to enable more robust grasping techniques. The flexion and extension of the InMoov hand digits are consistent without any form of control feedback, however, the digits displayed massive overshoot in trajectory. An open-loop control method is developed to smoothen artificial hand digit trajectory while reducing trajectory overshoot and transient vibrations. By using this method, the prosthesis digit trajectory is tested against the trajectory of the human finger, and found to be more consistent. Future works may focus on closed-loop control which rely upon more sensors to control the grasping of an object.