Algorithms for Human Presence Detection Based on Capacitive Sensor Data Processing

U okviru ovog rada obrađena je problematika detekcije osoba i objekata kapacitivnom metodom i algoritmi obrade informacija prikupljenih posebno projektiranim kapacitivnim senzorom i pratećom mjernom instrumentacijom. Opisana je teorijska podloga na kojoj se temelji dizajn kapacitivnog senzora i za specifični problem detekcije odabrani su odgovarajući algoritmi strojnog učenja za obradu mjernih podataka. Pomoću Robin Hood solvera provedeno je optimiranje mjernog sustava te su prikupljeni skupovi podataka za ispitivanje performansi algoritama strojnog učenja. Predložen je jednostavni mjerni sustav za detekciju prisutstva osobe i usporedbu korištenih metoda strojnog učenja. Razvijena je programska potpora koja implementira odabrane metode strojnog učenja. Skupovi podataka za učenje dobiveni putem Robin Hood simulacija pohranjeni su u odgovarajućem obliku i predstavljaju temelj za daljnja istraživanja mogućnosti primjene predloženog pristupa.This thesis elaborates the problem of people and object detection using capactive method, and algorithms for processing information acquired by custom designed capactive sensor and measurement instrumentation. Background theory for capacitive sensor design for object and people detection was presented, and machine learning algorithms for measurement data processing suitable for specific detection problem were selected. Robin Hood electrostatics solver was used in a process of measurement system optmization, and datasets for testing the performance of machine learning algorithms were generated. The simple measurement system for people detection and comparison of data processing methods was proposed. The software that implements chosen machine learning algorithms was developed. Datasets obtained by Robin Hood simulations were stored in appropriate data format, and they can serve as a basis for future research of possibilities of the proposed approach