The project consists of the following five work packages (WP) with different Tasks (T):
WP1: Design of DNNs for estimation of latent spaces of MUAPs. The objective of this WP is to define different architectures of DNNs for the estimation of the latent space of MUAPs and to test their potential for time-efficient tracking of MUAP shapes in fatiguing and dynamic contractions.
- T1.1 Preparation of experimental MUAP libraries
- T1.2 Design of DNNs for estimation of latent spaces of MUAPs
- T1.3 Quantification of MUAPs changes in fatiguing conditions
- T1.4 Quantification of MUAP changes in dynamic conditions
- T1.5 Prediction of MUAP changes during fatiguing and dynamic contractions
WP2: Design of DNNs for estimation of latent spaces of HDEMG. This WP will extend the work in WP1 on individual MUs to the global level of muscle excitation directly from HDEMG (without its decomposition to contributions of individual MUs).
- T2.1 Improvement of HDEMG simulations for learning sets in dynamic and fatiguing conditions
- T2.2 Design of DNNs for estimation of muscle excitation in latent spaces of HDEMG
- T2.3 Quantification of muscle excitation in latent spaces of HDEMG in fatiguing conditions
- T2.4 Quantification of muscle excitation in latent spaces of HDEMG in dynamic conditions
WP3: Quantification of variability in latent spaces of DNNs and translational learning. We will test the variability of MUAP and HDEMG properties in latent spaces of DNNs and analyse the capabilities and exploitation potential of their translational learning. We will collect large sets of HDEMG signals from the upper and lower extremities.
- T3.1 Intra-personal clustering of MUs in the latent spaces of MUAPs and translational learning in individual persons
- T3.2 Inter-personal clustering of MUs in the latent spaces of MUAPs and translational learning across different persons
- T3.3 Intra-personal clustering of muscle excitation profiles and translational learning in individual persons
- T3.4 Inter-personal clustering of muscle excitation profiles and translational learning across different persons
WP4: Demonstration and impact of the project results. This WP will demonstrate the wide exploitation potential of the methodology designed and data acquired in WP1-WP3. We have identified the following four demonstrators:
- T4.1 Improved MU identification in dynamic and fatiguing conditions
- T4.2 HDEMG channel replacement/repairing and AI-based increase of the number of recorded channels.
- T4.3 Generative AI for HDEMG signals
- T4.4 Revision of muscle synergy estimation methods and reassessment of muscle control synergies.
WP5: Project and data management, exploitation and dissemination. The objectives of this WP are the management of the project and internal Intellectual Property Rights (IPR), dissemination and communication activities and scientific and socio-economic exploitation of the results.
- T5.1 Project management with risk analysis
- T5.2 Data management
- T5.3 Measures to maximise the impact
- T5.4 Dissemination and communication of project results
The following data and knowledge management principles will be enforced during the project implementation:
- Open policy
- Knowledge management
- Data management
- Dissemination
- IP management
The project coordinator (Dr. A. Holobar, UM FEECS) will supervise all processes and development activities and ensure that the partners will follow existing quality rules, regulations and ethical requirements, with the support of the external ethics advisors (when required), and that the work performed is fully compliant with the legal and ethical requirements. This includes a) risk monitoring, b) risk assessment, c) coordination of contingencies.
