Thesis Project Form

For individuals with limb loss, the residual muscles of the stump are still able to contract, generating useful information that can be recorded using appropriate sensors. The most widespread signals used in prosthesis control is electromyography (EMG), primarily due to its non-invasive nature and ease of acquisition, as EMG sensors are applied directly to the skin surface. However, despite these advantages, EMG signals are inherently non-stationary, which poses challenges for robust prosthetic control in real-world settings. This variability arises from several physiological and environmental factors, including limb position, muscle fatigue, skin conditions such as sweating, and electrode displacement [1]. These fluctuations alter the properties of the EMG signal both within and across sessions, leading to changes in the data distribution that can degrade control performance and result in classification errors. Over time, such inconsistencies may contribute to user dissatisfaction and ultimately to prosthesis abandonment. This variability still represents a great challenge [2], and no universal solution has been presented to both quantify and eliminate problems such as the limb position effect [3]. The main goal is therefore to develop a novel approach to reduce such effect, starting from the experimental procedure to the design of a multimodal platform with the ultimate goal of being fully integrable in the prosthesis.

The primary objective of this thesis is to investigate the effect of limb position on high-density EMG (HD-EMG) signals and to develop an algorithm capable of maintaining robust performance despite this variability. To achieve this, the developments should exploit the multimodal nature of the BioInterNect platform, combining HD-EMG data with information from inertial measurement units (IMUs) to enhance control robustness.
The student’s activities will include:
• Experiment Design: Design of experimental protocol and wearable system to effectively collect data from subjects.
• Data Acquisition: Collect HD-EMG and IMU data to evaluate the effect of limb position on EMG signals. Data will be gathered both in a controlled laboratory environment and in a realistic setup that includes ADL-like tasks.
• Algorithm Development: Design and implement algorithms to mitigate the variability induced by limb position. The development will begin with offline analysis and progress to an online implementation, with the ultimate goal of integration into the embedded control platform of the prosthetic device.

[1] I. Kyranou, S. Vijayakumar, and M. S. Erden, “Causes of performance degradation in non-invasive electromyographic pattern recognition in upper limb prostheses,” Front. Neurorobotics, vol. 12, p. 58, 2018.
[2] M. R. Masters, R. J. Smith, A. B. Soares, and N. V. Thakor, ‘Towards better understanding and reducing the effect of limb position on myoelectric upper-limb prostheses’, in 2014 36th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Chicago, IL: IEEE, Aug. 2014, pp. 2577–2580. doi: 10.1109/EMBC.2014.6944149.
[3] E. Campbell, A. Phinyomark, and E. Scheme, ‘Current Trends and Confounding Factors in Myoelectric Control: Limb Position and Contraction Intensity’, Sensors, vol. 20, no. 6, p. 1613, Mar. 2020, doi: 10.3390/s20061613.

1. Programming skills: Matlab/Python, Programming of CAN-BUS/Wi Fi interface;
2. Practical skills: Soldering and Electronic basic techniques;
3. Design, implementation, and execution of experiments.
4. Analysis of biosignal data and feature extraction;
5. Development of multi-modal ML control algorithms for a multi-joint prosthetic device.