Development and implementation of computational intelligence architectures for self-interference cancellation in bidirectional radio communications (T4)

Research challenges / Novelty / Innovation
• Crosstalk cancellation in the analog domain using off-the-shelf components and FPGA implementation of the control algorithm
• FPGA implementation of crosstalk cancellation algorithms in the digital domain
• Hardware architectures/accelerators for implementing computational intelligence algorithms
• System integration and characterization of the bidirectional receiver

Research results:
• 1 prototype for implementing in-band bidirectional communication based on an SDR platform assembled from discrete components and FPGA/transceiver development boards (such as AD9361), respectively the characterization of the prototype

Innovation:
• Implementation of in-band bidirectional communication
• Application of echo cancellation and blind signal separation algorithms to mitigate crosstalk in the communication channel
• Signal processing architectures and neural network designs targeting implementation on FPGA or even ASIC

Implementation on FPGA of a cardiac monitoring solution integrating AI and local ML inference

Objective: Implementation on FPGA of a cardiac monitoring solution integrating artificial intelligence and local ML inference. Real-time, low-latency and energy-efficient solutions are considered for monitoring physiological data and identifying pathological patterns in ECG, suitable for edge-AI applications and wearable devices.

Research challenges / Novelty / Innovation
• ECG signal preprocessing and extraction of relevant features
• Development and validation of AI/ML models for cardiac monitoring and assessment
• Optimization of AI/ML models for FPGA implementation
• FPGA implementation and integration for interoperability
• Testing and validation under controlled conditions

Research results:
• Hardware prototype validated in laboratory

Innovation:
• Integration of complex AI models into optimized embedded architectures

Implementation of multimodal biometric recognition systems based on electrophysiological signals on FPGA

Objectives: implementation of multimodal biometric recognition systems based on electrophysiological signals: ECG, PPG, PCG, using artificial intelligence techniques implemented on FPGA

Research challenges / Novelty / Innovation
• Processing biomedical signals and extracting relevant features
• Design of AI algorithms for biometric recognition
• AI optimization for hardware implementation
• FPGA implementation and system integration
• Testing and validation under controlled conditions

Research results:
• Functional integrated architecture

Innovation:
• Full AI implementation on FPGA for biometric recognition from biophysiological signals