0566001 20230316110215.220221229235959.9 Prague České vysoké učení technické v Praze 2022 182 s. P RLS algorithms FIR filters hypothesis testing Bayesian approach FPGA parallel processing pipelining ultrasound hand detection cav_un_auth*0330517 Likhonina Raissa UTIA-B Zpracování signálů Department of Signal Processing ZS ZS Department of Signal Processing CZ Ústav teorie informace a automatizace AV ČR, v. v. i. http://library.utia.cas.cz/separaty/2022/ZS/likhonina-0566001.pdf The thesis is devoted to fast Bayesian algorithms, more precisely to the QRD RLS Lattice algorithm combined with hypothesis testing and applied to hand detection problem solution based on ultrasound technology. Due to the proposed structure of regression models and the offered approach to hypothesis testing in the work, the algorithm under consideration is able to solve the problem of noise cancellation and additionally to compute the distance between the hand and the device; thus, potentially enabling to identify simple gestures. Further, the algorithm was implemented in parallel on the HW platform of Xilinx Zynq Ultrascale+ device with a quad-core ARM Cortex A53 processor and FPGA programmable logic and proved to function reliably and accurately in real time using real data from an ultrasound microphone. The work contains an investigation of the state of the art in the corresponding field and gives the theoretical background necessary for the development and modification of the algorithm to fulfill the goals of the thesis. The thesis also includes thorough description of experiments and an analysis of the results including those from simulation and from computation using real ultrasound data both in the MATLAB R2019b environment and on the HW platform of Xilinx Zynq Ultrascale+. BC 10000 10200 10201 2023 Ph.D. ČVUT, Fakulta dopravní Prague 2022 21.10.2022 RVO:67985556 https://hdl.handle.net/11104/0337922 S 2022 2022 Prague České vysoké učení technické v Praze