Toward Stable and Reliable Lower Limb Prosthetics Control with Signals Recorded from Muscles

Doctoral thesis, 2025

Prosthetic devices are essential in enhancing mobility and functionality for individuals with amputations, enabling them to perform daily activities with improved independence and ease. The effectiveness of these prosthetic devices depends significantly on their design, functionality, and the user's ability to intuitively control it and rely on it. In pursuit of enhancing control, our research focused on the integration of electromyography (EMG) signals into prosthetic control as a means to detect movement intention of the users. EMG signals offer a promising avenue for developing more natural and intuitive prosthetic systems. Although this technology has successfully improved the functionality of prosthetic arms, its application in prosthetic legs has been less extensively explored. 
This research aimed to extend the use of EMG technology to lower limb prosthetics, drawing from the established successes in upper limb applications. While the use of EMG for lower limb prosthetics has been investigated in prior studies, it remains less extensively explored and adopted compared to upper limb applications. To this end, we developed an open-source software framework for acquiring and processing biological data, such as electromyography (EMG), and non-biological data, including  inertial measurement units (IMUs). This framework aims to foster collaboration and drive innovation within the global scientific community by encouraging researchers to actively develop, compare, and enhance algorithms, thereby accelerating progress in prosthetic technology. We conducted a benchmark test using our open access dataset from 21 able-bodied individuals to validate the platform's effectiveness.  
Building on this validation, we tested the system with individuals living with limb loss, the next critical step in achieving robust and reactive control of prosthetic legs. Furthermore, to address the challenges associated with traditional socket-based systems for EMG-controlled prosthetics—such as signal instability and user discomfort—we recorded EMG signals from individuals with osseointegration. Osseointegration eliminates the need for a socket by providing a direct connection between the prosthetic and the skeletal structure, resulting in more stable electrode placement and reducing motion artifacts caused by shifting soft tissues. This improves EMG signal quality and consistency, allowing our algorithms to interpret more accurately the users' intended movements. To further enhance the accuracy and reliability of movement predictions, we refined our intention detection algorithms by incorporating post-processing techniques specifically designed to filter out low-confidence predictions from the EMG and IMU data, reducing the risk of incorrect intention detection and preventing unintended prosthetic movements.  
We also explored the integration of neural signals to enhance the responsiveness of prosthetic devices, aiming for more intuitive and seamless user interactions. In addition, the final phase of this research focused on the development of a clinical rehabilitation protocol aimed at users of active prosthetic legs and neuromusculoskeletal interfaces. These initial efforts represent the foundational steps for broader adoption of EMG-based control systems in lower-limb prosthetics, with the potential to substantially improve users' quality of life.