Klasifikacija Wi-Fi signalov in merjenje obiska po regijah

ŽUPEC, NEJC (author), Ciglarič, Mojca (mentor)

Vse več je mobilnih naprav, ki oddajajo signale Wi-Fi. Te lahko zaznamo z dostopnimi točkami, posledično pa lahko merimo število obiskovalcev v bližini dostopnih točk. Pogosto nas zanima število obiskovalcev samo za določen prostor oz. regijo. Zato smo v okviru magistrske naloge raziskali metode in produkte, s katerimi je možno določiti lokacijo izvora signala. Ker nas zanima le regija, smo problem določanja lokacije transformirali v problem klasifikacije signala Wi-Fi v regije. Po zgledu obstoječih metod za določanje lokacije (metoda najmočnejša dostopne točke, trilateracija in sistem RADAR), smo razvili tri nove metode. Da bi lahko te metode ovrednotili, smo na Fakulteti za računalništvo in informatiko postavili sistem za zajem signalov Wi-Fi in z njim dva meseca zajemali podatke. S pomočjo strojnega učenja smo razvili metodo, ki v 85,1 % pravilno napove regijo izvora signala Wi-Fi. V primeru, da so regije med seboj ločene s stenami, pa klasifikacijska natančnost znaša več kot 93 %.

• Language: Slovenian
• Keywords: Wi-Fi, klasifikacija, SVM, strojno učenje, RADAR, trilateracija, določanje lokacije
• Work type: Master's thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2015
• PID: 20.500.12556/RUL-30861
• Publication date in RUL: 07.07.2015

ŽUPEC, NEJC, 2015, Klasifikacija Wi-Fi signalov in merjenje obiska po regijah [online]. Master’s thesis. [Accessed 25 April 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=30861

Secondary language

• Language: English
• Title: Wi-Fi signal classification and visitor counting by region

Abstract

More and more mobile devices transmit Wi-Fi signals which can be detected by access points. Consequently the number of visitors near access points can be measured. Often we are interested in the number of visitors only for a certain region. Therefore, in the context of master's thesis, methods and products for indoor localization were studied. We transformed the localization problem into classification of Wi-Fi signals problem. Based on existing methods (nearest base station, trilateration and RADAR) we developed three new methods. In order to evaluate these methods the sytem for capturing Wi-Fi signals was set up at the Faculty of Computer and Information Science in Ljubljana. The data was being collected for two months. Based on machine learning algorithms we developed a new method, which correctly predicts the region in 85,1 % cases. If regions are separated by walls, the classification accuracy is more than 93 %.

• Keywords: Wi-Fi, classification, SVM, machine learning, RADAR, trilateration, localization