Between August and September 2019, I was fortunate enough to be invited for a three-week internship at RITMO supervised by Prof Kyrre Glette, to work with surface electromyography (sEMG) (i.e. muscle activation). sEMG provides an intuitive and non-intrusive interface from which one can control robotic devices by performing different hand/wrist gestures. The gap between the user’s intention and the muscle signal is bridged by using a classifier. However, four main dynamic factors influence the signal activity in such a way that the learned mapping by the classifier can become erroneous, severely limiting the real-time applicability of such an interface.
These four main factors are:
- Limb Position
- Contraction Intensity
- Electrode Shift
- Inter-day recording
One key component of my PhD research is to develop a self-adapting algorithm able to cope with these factors. During my stay at RITMO, we worked on developing a virtual reality-based experiment to study and collect data which will enable me to test theses algorithms in as close to a real setting as possible.
Generally, the limb position is studied by asking the participant to perform the different gestures at specific, pre-determined limb positions (e.g. forearm parallel to the floor and 90-degree elbow bent, forearm 45-degree from the floor and 120-degree elbow bent). However, this method of discrete positioning severely limits the amount of limb position that can be studied. Instead, the virtual reality software also leverages the leap motion camera to track the actual limb position of the participant in real-time. The limb position is modulated in real-time by asking the participant to reach a random position while performing the gestures.
Contraction intensity was requested using a color mapping (blue=low intensity, yellow=medium intensity, red=high intensity) that showed on the ring and the hand model (see Figure 1). The color of the participant’s virtual hand changed according to the detected contraction intensity, offering intuitive and real-time feedback.
Finally, for each participant, the experiment is performed for 14 days and the participants are required to place the sEMG-based armband themselves. These last two points ensure that the recorded dataset will include the electrode shift and inter-day recording factors.
The emphasis on interdisciplinarity at RITMO and the experience of the associated researchers in rhythm, time, motion and surface electromyography made the center a place of excellence to develop this software. A video showing the developed software in action is available here
My time at RITMO was an extremely positive and enriching experience. The interdisciplinary nature built at the core of the center ensured that I was able to collaborate with researchers that could understand the problems I had and offer a completely new perspective on how to solve them. The center also has state of the art motion capture facilities and offered multiple communal activities to connect with the members of the center more easily. Even though I was there only three weeks I felt extremely welcomed and I hope to back soon to collaborate on new projects. The free hot chocolate was a bonus!