A voice-controlled bionic hand performed preset gestures, repeated its finger movements and carried out basic grasp tests in an evaluation of an engineered prototype. The reported tests show the system translated fixed commands into coordinated finger actions, but they do not show how it would work for a person.
The study focused on a simplified actuation mechanism for coordinated finger movement. Its evidence came from functional and mechanical observations of the hand, including synchronized movement, preset gestures, sequential flexion, voice-response checks and tests with soft and rigid objects.
A compact control chain
The control chain combined voice recognition, Arduino control units, Bluetooth communication and servo actuation. Five free-standing servo motors drove finger movement through a tendon system, while the hand’s components were fabricated by additive manufacturing with Polylactic Acid, or PLA.
The system used open-loop control. In practical terms, predefined commands directly triggered actuator actions without real-time feedback from the hand. Servo calibration assigned approximately 0 degrees to the closed state, 45 degrees to partial flexion and 90 degrees to the open state, with about plus or minus 5 degrees of tolerance.
Preset actions, conditional grip
The preset gesture logic supported both synchronized and selective movements. The reported set included an open hand, a fist, a pinch and a half-grip, allowing several fingers to move together or selected fingers to move for a particular action.
When flexion was performed sequentially, all fingers reached the fully flexed position in about 7 to 8 seconds. Across several actuation cycles, the motion was described as repetitive and constant, with no apparent loss of tension or control precision.
Voice inputs were linked to the open-hand, fist, pinch and point movements. The report says the identified commands produced their associated actions with no observable delay, although it does not give a measured latency or a command-accuracy rate.
The grasp tests used soft and rigid objects approximately 1 to 5 cm in size and assessed grasp stability under different conditions. Smooth surfaces produced minor slipping, while changes in object thickness led to either a poor grip or one that was too tight.
The control system did not receive real-time feedback, and the tests did not demonstrate a way for the hand to adjust its grip as surface and thickness changed.
What the tests leave open
The authors characterize the prototype as practical, modular, highly repeatable and simple to implement. The evidence for that assessment is limited to the reported prototype observations, which do not include a quantitative repeatability measure or a count of the actuation cycles used.
The same boundary applies to the voice interface. The report describes responses to identified voice inputs, but does not establish command accuracy or robustness across different users and operating conditions.
Fatigue behavior and long-term durability were not systematically evaluated. Possible wear, tendon elongation and changes in tension distribution under extended use therefore remain open questions.
Evaluation was restricted to functional and mechanical performance. No human-user validation or ergonomic evaluation was conducted, leaving safety, comfort and usability untested.
The authors propose human-user evaluation and closed-loop force sensing as future steps. In the present open-loop tests, smooth objects could slip and thicker objects could be held poorly or too tightly.
As an engineering concept, the work offers a modular voice-to-motion setup for exploring tendon-driven hand control. As evidence for a device used by people, however, it remains limited to prototype performance rather than human-user validation.
Paper data and sources
Original title: Development of a Voice-Controlled Tendon-Driven Bionic Hand
Authors: Urja Kohli, Shagata Chanda, Kritika Gandhi et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text