Design and Functional Evaluation of a Dexterous Myoelectric Hand Prosthesis With Biomimetic Tactile Sensor

This paper presents the design, tactile sensor, characterization, and control system of a new dexterous myoelectric hand prosthesis to overcome the limitations of state-of-the-art myoelectric prostheses (e.g., limited functionality, controllability, and sensory feedback). Our dexterous myoelectric hand allows independent finger movement and thumb abduction/adduction, with a motor for each finger and an additional motor for the thumb (i.e., six total motors). Each fingertip has a biomimetic tactile sensor with 13 tactile units, each of which can detect normal and tangential forces. The hand controller uses an electromyography pattern recognition controller and a tactile sensor feedback-based grasping controller to automatically and dynamically adjust the finger grasp force to prevent objects from slipping. This closed-loop controller structure will allow users to safely and effectively grasp complex objects with varying densities and shapes. In addition, the electronic hardware is integrated into the hand, and the pattern recognition controller can be implemented in the hand embedded system.

I. Introduction

State-of-the-art myoelectric hand prostheses, such as i-limb [1] and Vincent Hand [2], allow the user to accomplish more tasks compared to commercial hand grippers. However, myoelectric hand prostheses still have limitations, such as reduced functionality, limited controllability, lack of sensory feedback, and cosmesis, as well as wearing discomfort [3]. To overcome the low functionality of commercial grippers, many multifingered prosthetic hands have been developed (e.g., HIT hand [4], CyberHand [5], SmartHand [6], RTRII Hand [7], SDM hand [8], UBH-IV [9], ACT Hand [10], RIC arm [11], Multigrasp Hand Prosthesis [12], and Vanderbilt hand [13]). A major trend of these prototypes is the addition of more degrees of freedom (DOFs) to improve the prosthetic hand dexterity. In fact, some commercial multifingered prosthetic hands, such as i-limb [1], Vincent Hand [2], and Bebionic v2 [14], perform well because they have more than five DOFs. The result is that a multifunctional prosthetic is capable of different grasping patterns, such as precision, power, and lateral grips. Despite the technical achievement of all these advantaged prosthetic hands, their functionality is still limited. Some common tasks, such as picking up delicate objects, are still big challenges and additional development are still needed to address these challenges. Picking up a delicate objects need very fine control of the prosthetic hand and could fail if any of the issues could not addressed correctly: 1) choose and maintain a correct grasp pattern; 2) sense the contact force between the hand and the object; 3) adjust grasp force in real-time to prevent delicate objects from slipping and being broken; and 4) generate a well packaged hand platform, of which rely on external components is minimized.