Radiomyography for Real-Time Control of Upper-Limb Prostheses: A Wearable RF Sensing Approach
Translated title
Radiomyography for Real-Time Control of Upper-Limb Prostheses
Author
Ladekjær-Mikkelsen, Jens-Ulrik
Term
4. semester
Education
Publication year
2026
Submitted on
2026-06-03
Abstract
This thesis describes the development and testing of a portable control system for hand prostheses based on RMG (radio-frequency muscle sensing). The goal is to make upper-limb prosthesis control more intuitive for users. Unlike the conventional technique sEMG (surface electromyography), which uses skin-contact electrodes, RMG relies on radio-frequency (RF) sensing to detect muscle activity. This allows muscle signals to be measured without direct contact with the skin and makes it possible to sense both superficial and deeper muscles. The proposed system includes custom-designed patch antennas mounted on a wearable forearm brace, a LiteVNA device for RF signal acquisition, a Raspberry Pi 5 for embedded data processing, and a Linkerbot L7 robotic hand that performs the movements. To study how the antennas transmit signals in the arm and to assess safety, electromagnetic simulations were carried out in CST Microwave Studio. These simulations used voxel-based anatomical forearm models (built from many small volume elements) to evaluate signal propagation, coupling, and safety aspects of the antenna system. Measurements from the antenna system are represented as S-parameters, which describe how the RF signal is altered. These data were processed using machine learning methods to classify different hand gestures (gesture recognition). In controlled experiments, the system achieved gesture classification accuracies of up to 98.9%, while tests across different recording sessions (measured at different times) reached around 82% accuracy. Real-time experiments showed that, with a total processing cycle time of approximately 315 ms, the system could successfully control multiple grasp types in the robotic hand. Overall, the results demonstrate that a portable RMG-based prosthetic control system is technically feasible and that RMG has strong potential as a future muscle sensing method for advanced hand prosthesis applications.
Dette speciale beskriver udviklingen og afprøvningen af et bærbart styresystem til håndproteser, som bygger på såkaldt RMG (radiofrekvens-muskelsensing). Målet er at gøre styringen af overekstremitetsproteser mere intuitiv for brugeren. I modsætning til den traditionelle teknik sEMG (overflade-elektromyografi), som bruger elektroder på huden, anvender RMG radiobølger (RF-sensing) til at registrere muskelaktivitet. Det gør det muligt at måle muskler uden direkte hudkontakt og at opfange signaler både fra overfladiske og dybereliggende muskler. Det foreslåede system består af specialdesignede patchantenner monteret på en bærbar skinne til underarmen, et LiteVNA-instrument til at indsamle RF-signaler, en Raspberry Pi 5 til indlejret databehandling samt en Linkerbot L7 robot-hånd, der udfører bevægelserne. For at undersøge, hvordan antennerne udbreder signaler i armen, og om systemet er sikkert, blev der lavet elektromagnetiske simuleringer i programmet CST Microwave Studio. Disse simuleringer brugte anatomiske modeller af underarmen opbygget af små volumenelementer (voxel-baserede modeller) for at vurdere signaludbredelse, kobling og sikkerhed. Målingerne fra antennesystemet bliver repræsenteret som såkaldte S-parametre, der beskriver, hvordan RF-signalet ændres. Disse data blev analyseret med maskinlæring, så systemet kunne genkende forskellige håndbevægelser (gestusklassifikation). I kontrollerede forsøg opnåede systemet en nøjagtighed på op til 98,9 % i klassifikationen af håndbevægelser, mens test på tværs af forskellige sessions (målt på forskellige tidspunkter) gav omkring 82 % nøjagtighed. Realtidsforsøg viste, at systemet med en samlet behandlingstid på cirka 315 ms pr. cyklus kunne styre flere forskellige typer greb i robot-hånden. Samlet set viser resultaterne, at et bærbart RMG-baseret prostesestyresystem er teknisk muligt, og at teknologien har potentiale som en fremtidig metode til at måle muskelaktivitet i avancerede håndproteser.
[This abstract has been rewritten with the help of AI based on the project's original abstract]
Keywords
RMG ; RF ; Muscle sesing ; Prothesis