Project Partners
Signal Processing
- Real-time, multi-channel and multi-modal speech enhancement
- Integrate neuroscientific models into common neural networks
- Compare bio-inspired vs. mathematically optimized models
- What works better ("Is it worth informing neural nets from human brains?")
- Perceptual losses and metrics
Human-Computer Interaction
- Focus on usability and real-word applicability
- Smart Glasses
- Make use of Compantion App / VR testbed for testing
- Requirement analysis
- With potential users and corporations (Bionics, Fielmann, Nuance)
- Keeping requirements minimal
- Integrating gaze tracking and multimodal feedback
- Configurable settings via user interface
- Formative and summative evaluations
- Oscillatory systems
- Theta-Gamma waves as for dynamic gain on phoneme and syllable attention
- Phase-Locking (Input signal <=> Spiking/Oscillations)
- Network Topology
- [[Global Neuronal Workspace (GNW)]] as inspiration
- Test parameter variation of
- network density
- interlinking (dense vs. sparse)
- hierarchical modularity (modal inputs via peripheral modules, fused in GNW)
- Interplay of topology and oscillations
- [[Predictive Coding]]
- Infer brain mechanisms from what works well with AI model
- Requirement and feedback from UKE and Hospital zum Heiligen Geist
- Ethical concerns and data management