Design and Prototyping Soft Fingers for a Hand Prosthesis with the Aim of Predicting the Gripping Force Using LSTM

Amputation, which has long occupied the human mind, has tremendous negative effects on different aspects of amputees’ lives. Hands, amongst the principal organs in the human body for communication, are capable of performing copious tasks. Consequently, it is essential to identify how this organ and its various components function, in order to find a suitable way to design and build a system that can be an applicable alternative for them. Designing a cost-effective prosthesis that not only has a fast and reliable performance, but also has an adequate range of motion is no easy task. Since grasping and manipulating objects are the major functions of the hands and force feedback can enhance prosthetic hands’ grasping abilities, this study presents a low-cost mechanical design for soft fingers with the aim of measuring the applied pressure on the fingertips. These fingers are attached to a prosthetic hand to gain pressure feedbacks, which can play a key role in the gripping control of the prosthesis. Finger pressure and gripping force during multiple precision-grips are recorded, and gripping force is adequately predicted from fingertip pressures using a long short-term memory neural network with an R-squared value and a mean squared error of 0.9604 and 0.00085 for normalized data, respectively.