Most prosthetic hands still face a fundamental challenge: every amputee has unique needs, yet most devices rely on a one-size-fits-all design. As a result, achieving natural, intuitive control remains difficult, often requiring users to spend significant time learning and adapting.
Even with today’s sophisticated prosthetic technology, many users still struggle to control their devices. Muscle signals can fluctuate because of sweat, changes in skin condition, or routine movements, making it difficult for the prosthesis to consistently interpret the wearer’s intentions. This disconnect can be frustrating and may even cause some users to stop using their prosthetic hand altogether.
Although researchers have improved the way prosthetic systems decode muscle activity, a key obstacle persists: these signals remain inconsistent and difficult to convert into smooth, intuitive movements.
To overcome this limitation, Erik Engeberg, Ph.D., is developing a different approach that prioritizes personalization over one-size-fits-all designs. A professor in Florida Atlantic University’s College of Engineering and Computer Science, Engeberg holds appointments in the Departments of Ocean and Mechanical Engineering and Biomedical Engineering. He also belongs to the FAU Stiles-Nicholson Brain Institute and the FAU Center for Complex Systems, where he develops prosthetic systems that adapt to each user’s unique characteristics.
A Tailored Sleeve Interprets User Intent
The process starts with a 3D scan of the user’s residual limb, which engineers use to design a custom 3D-printed wearable sleeve fitted with soft, flexible magnetic sensors. These sensors rest comfortably on the skin and detect subtle shifts in muscle shape and pressure when the user tries to move their hand or wrist, enabling real-time interpretation of intent.
Engineers fully personalize the system: they adjust the sensor layout to include either 18 or 24 modules depending on limb size and anatomy. Engineers also pair it with an AI model trained specifically for each individual, which learns that person’s unique muscle activity patterns instead of relying on generic datasets.
Consistent Signals During Repeated Use
In tests involving 10 participants, including three upper-limb amputees, the system was able to recognize 19 hand and wrist gestures in real time and convert user intent into control of a dexterous robotic hand. The findings, published in IEEE Transactions on Neural Systems and Rehabilitation Engineering, indicate that performance remained reliable and consistent across repeated trials.
To evaluate long-term durability, the researchers subjected the system to more than 7,500 robotic force cycles over several hours while carefully monitoring sensor responses. The results demonstrated a strong and stable correlation between applied force and output, with accurate pressure detection and no noticeable decline in performance.
After thousands of repeated cycles, the signals remained clear and stable, showing strong separation between noise and meaningful data and only slight variation over time. Overall, the sensors exhibited no significant drift or degradation, preserving the accuracy, repeatability, and responsiveness required for practical prosthetic use.
Senior author Engeberg said, “We cannot standardize prosthetic control because every person has a unique movement pattern shaped by their anatomy, injury background, and the way their remaining muscles operate.” “For these systems to be effective in everyday use, they must be tailored to the individual. By integrating 3D-printed wearable sensors with personalized AI models, we are getting closer to prosthetics that respond naturally and instantly to a user’s intent, instead of requiring users to adapt to the device’s constraints.”
There Is No Single Sensor Configuration that Fits Everyone
The results also revealed that there is no universal sensor configuration that works best for everyone. Some participants reached higher accuracy with fewer sensors, while others needed a greater number, with the optimal setup depending on individual anatomy, injury history, and the condition of the remaining muscles. In several cases, participants achieved over 90% accuracy across multiple gestures only when engineers customized the sensor arrangement to match each user’s residual muscle patterns.
“Our findings show that prosthetic performance strongly depends on how well sensor placement and density are adapted to the individual,” said Engeberg. “This points toward a future where prosthetists can adjust sensor configurations much like a medical prescription, optimizing both comfort and functionality for each user.”
The study also generated a shared dataset from both amputee and non-amputee participants, offering a useful resource for further research in the field.
The Extent of the Unmet Need
“This research addresses a very practical goal: directly improving people’s quality of life,” said Stella Batalama, Ph.D., dean of the College of Engineering and Computer Science. “When engineering advances are aligned with real-world needs—especially for individuals who rely on prosthetic devices for independence—the impact extends well beyond the laboratory. It is about restoring function, rebuilding confidence, and enabling people to interact with the world more naturally.”
In the United States, about 2.1 million people are living with limb loss, with roughly 185,000 amputations performed each year. Worldwide, more than 50 million people are affected, and this number is expected to rise due to diabetes, vascular disease, trauma, and conflict-related injuries. Restoring upper-limb function is particularly difficult because of the intricate complexity of natural hand and finger movements.
The study’s co-author is Wen-Yu “Marty” Cheng, a graduate student and Ph.D. candidate in Florida Atlantic University’s College of Engineering and Computer Science.
Read the original article on: Tech Xplore
Read more: A new smart bicycle can detect when a rider intends to turn and when they might be losing balance