Partial hand test fixture


Improving Control of Robotic Hand Prostheses


Partial-hand amputation is the most common type of upper limb amputation, affecting an estimated 500,000 individuals in the United States as of 2005. Between 1988 and 1999, more than 150,000 Americans had amputations through the palm of the hand (1.5%), of the thumb (18%), or of one or more digits. The most frequent cause of partial-hand amputation is trauma, often a result of work– or leisure activity–related accidents or combat injury. Thus, the majority of partial-hand amputations occur in relatively young, active individuals.

Losing fingers or the thumb impairs the ability to make functional hand-grasps, which are essential for eating and dressing, leisure activities, and job-related tasks. More than half of all partial-hand amputees are unable to return to previous employment, and most must modify their work duties because their prosthesis does not provide sufficient function. In fact, partial-hand amputees often perceive themselves to have a higher level of disability than individuals with arm amputations at the elbow or shoulder level.

New, powered prostheses for individuals with partial-hand amputation are becoming available and can perform many of the functions of an intact hand; however, controlling these devices is challenging. Typically, users control these prostheses using electrical (electromyographic or EMG) signals generated by muscles in the hand (the intrinsic hand muscles). However, the intrinsic muscles are small and are often damaged by the amputation, so obtaining usable EMG signals can be difficult.

Muscles in the forearm (called extrinsic hand muscles) also control movement of the hand and fingers; however, these muscles lie close to the muscles that move the wrist, so using EMG from these muscles to control a prosthesis would mean the user has to keep their wrist still while they move their prosthesis. This would prevent them from using the wrist to position the hand, making many tasks harder and more awkward to perform.

These conventional control methods also only allow the user to control one type of hand grasp, so users must switch the prosthesis into different modes to perform other grasps.

A solution to these problems is to use pattern recognition technology to decode the EMG signals from extrinsic muscles. Any attempted movement of the hand or digits creates a pattern of EMG signals specific for that hand movement. A computer algorithm can ‘learn’ these patterns and then predict what movement the user intends to make, based on recorded EMG signal patterns. The user can thus control the prosthesis—and potentially perform a variety of hand movements—just by attempting to make the desired movement.

Our research goal was to develop a pattern recognition algorithm that enables accurate control of a partial hand prosthesis using EMG while still allowing the user to move their wrist. We found that using combined EMG from multiple muscles together with simulated wrist position information that predicts EMG changes in different wrist positions reduced the effect of wrist position and improved accuracy of the control system. In addition, we can accurately predict what the EMG would look like in different wrist position, so the user only has to train the pattern recognition system in one wrist position

In the second phase of the project, we compared the ability of individuals with partial hand amputations to perform a set of functional activities using our pattern recognition control system and a conventional control strategy. Four individuals with partial hand amputations were fit with prosthetic digits from Touch Bionics Inc.

All subjects could successfully use both conventional control and pattern recognition modes to control this prosthesis and complete assigned tasks, including a custom-designed ‘cubbies’ task (in which subjects moved an object between compartments in a 9-cube (3X3) storage unit) that requires wrist movement.

Using pattern recognition, all subjects were able to successfully complete the tasks using EMG from extrinsic muscles only and from a combination of extrinsic and intrinsic muscles, and some subjects were also able to complete all movements using only EMG from intrinsic muscles. Although performance of the pattern recognition system and direct control system during testing were equivalent during the tasks, the pattern recognition system allowed users to intuitively select the hand-grasp they desired, without requiring mode switching.

The control technology has been licensed by a start-up company, Coapt LLC, for commercial sales, and knowledge gained on fitting transradial amputees with pattern recognition control systems has also been transferred to Coapt.

Related Publications



Adewuyi A, Hargrove L, and Kuiken TA. Intrinsic and Extrinsic Hand Muscle EMG Improve Pattern Recognition Control and Mitigate the Effect of Wrist Motion. IEEE Transactions on Neural Systems and Rehabilitation Engineering. 24.4: 485-494, 2016.

Adewuyi A, Hargrove L, and Kuiken TA. An Analysis of Intrinsic and Extrinsic Hand Muscle EMG for Improved Pattern Recognition ControlIEEE Transactions on Neural Systems and Rehabilitation Engineering, (Epub May 6, 2015) 24(4): 485–494, 2016.

Earley EJ, Hargrove LJ, Kuiken TA. Dual Window Pattern Recognition Classifier for Improved Partial-Hand Prosthesis ControlFrontiers of Neuroscience, Neural Technology section, special topic: “Current challenges and new avenues in neural interfacing: from nanomaterials and microfabrication state-of-the-art, to advanced control-theoretical and signal-processing principles.” Frontiers in Neuroscience. 10:58. doi: 10.3389/fnins.2016.00058.

Adewuyi A, Hargrove LJ, and Kuiken TA. Evaluating EMG Feature and Classifier Selection for Application to Partial Hand Prosthesis ControlFrontiers in Neurorobotics 10:15, 2016.

Adewuyi AA, Hargrove LJ, Kuiken TA. Resolving the effect of wrist position on myoelectric pattern recognition control. Journal of NeuroEngineering and Rehabilitation 14:39, 2017.

Adewuyi A, Kuiken TA, and Hargrove LJ. A Comparison of Conventional and Pattern Recognition Myoelectric Control of Powered Partial-Hand Prostheses. In submission to PLos One.

Conference Presentations

Earley E, Adewuyi A, and Hargrove L. Optimizing Pattern Recognition-Based Control for Partial-Hand Prosthesis Application. IEEE conference in Engineering in Medicine and Biology, 26-30 August, 2014.

Adewuyi A, Hargrove L, and Kuiken TA. Towards Improving Partial Hand Prostheses: The Effects of Intrinsic Muscle EMG and Wrist Motion on Myoelectric Pattern Recognition. Myoelectric Control Symposium (MEC), Fredericton, New Brunswick, August 18-22, 2014.

Adenike A. EMG Feature Evaluation for Improved Myoelectric Control of Partial-Hand Prostheses. 7th International IEEE EMBS Neural Engineering Conference, April 22-24, 2015, Montpellier, France.

Adewuyi A. Pattern Recognition-Based Myoelectric Control of Partial Hand Prostheses, Summer School in Neuro Rehabilitation (SSNR), September 13-18, Valencia, Spain.

Earley EJ, Hargrove LJ. The Effect of Wrist Position and Hand-Grasp Pattern on Virtual Prosthesis Task Performance. IEEE RAS and EMBS International Conference on Biomedical Robots and Biomechatronics (BioRob); June 26-29, 2016, Singapore.

Adewuyi A, Hargrove LJ, Kuiken TA. Preliminary Functional Outcomes for Myoelectric Pattern Recognition-Based Control of Partial-Hand Prostheses. 38th International conference of the IEEE Engineering in Medicine and Biology Society, August 16-20, Orlando, FL.


The contents of this webpage were developed under a grant from the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR grant number 90RE5014-02-00). NIDILRR is a Center within the Administration for Community Living (ACL), Department of Health and Human Services (HHS). The contents of this webpage do not necessarily represent the policy of NIDILRR, ACL, HHS, and you should not assume endorsement by the Federal Government.

Save now, read later.