The Institute of Information Theory and Automation (UTIA) is a public non-university research institution which administratively falls under the Czech Academy of Sciences. UTIA conducts fundamental and applied research in computer science, signal and image processing, pattern recognition, system science, and control theory. In addition to its research activities, UTIA is engaged in undergraduate, graduate, and postgraduate education. It also publishes the journal Kybernetika and acts as a certified forensic expert institution. UTIA consists of eight scientific departments, computer centre, library, facility maintenance department and business administration department.
The Signal processing department, involved in Listen2Future, focuses on the research, development and implementation of advanced digital signal and image processing algorithms, mainly in the fields of telecom, audio processing and scene analysis. We build on our experience with the Bayesian approach to recursive identification of linear systems with time variable parameters.
Our target platforms are Field-Programmable Gate Arrays (FPGAs). We are focused mostly on embedded SoC solutions based on Xilinx Kintex, Zynq and ZynqULTRASCALE+ devices programmed by Xilinx Vivado and SDSoC tools. We also use Matlab/Simulink or OpenCV library to specify, model and verify algorithms which we subsequently convert and synthesize to FPGAs using HLS tools. That is why we also study features which result in fast execution, small memory footprint, small chip area and low power consumption. This is achieved through designing new DSP algorithms or modifying the existing DSP algorithms and by exploiting advanced architectural properties of FPGA circuits.
Currently we focus our development mainly around Zynq and ZynqULTRASCALE+ FPGA family boards from Xilinx and from Trenz (industrial and automotive grade ones).
Role in the project
UTIA will focuse on implementation of easy start FPGA platform with US transducer support. The platform will include capabilities for hardware acceleration of algorithms needed to implement pre-processing, edge AI, post processing and optionally also visualisation of results. UTIA will also collaborate with IMA and BUT on development and implementation of technology required for predictive maintenance and infrastructure monitoring. The results of our work will be demonstrated in UC9:” Infrastructure monitoring and maintenance optimization”.
Key contribution to the project
UTIA will develop SoC design support for simultaneous acoustic and video capture in order to provide match of acoustic and visual events useful for precise acoustic event localisation in infrastructure monitoring and maintenance optimization use case. It will also include developments of methods for calibration of audio and video-based measurements, post-processing and visualisation of detections.
Department of Signal Processing, Institute of Information Theory and Automation of the Czech Academy of Sciences
Pod Vodárenskou věží 4
180 00 - Praha 8
CZECH REPUBLIC
www.utia.cas.cz