Rekonstrukcija interaktivne instalacije s prepoznavo hoje

Godec, Miha (avtor), Bovcon, Narvika (mentor), Batagelj, Borut (komentor)

V okviru diplomske naloge je implementirana rekonstrukcija interaktivne instalacije z naslovom Matrica Koincidenca Zmenkarski klub iz leta 2006 (avtor: Srečo Dragan), ki temelji na podobnosti med ljudmi glede na podobnost njihove hoje. Delo je razdeljeno na dva glavna dela. V prvem je predstavljena metoda za analizo hoje, njene značilnosti in metoda za pridobivanje identifikacijskega vektorja osebe na podlagi le-teh, ki deluje s pomočjo nevronskih mrež. V drugem delu pa se osredotočimo na praktično uporabo in implementacijo uporabljene metode za rekonstrukcijo dela instalacije. V okviru instalacije smo obiskovalcem omogočili, da posnamejo svoje gibalne vzorce pri hoji in jih primerjajo z vzorci drugih obiskovalcev. Iskanje podobnosti temelji na metodi podpornih vektorjev, s katero smo na testnih podatkih dosegli 81-odstotno natančnost. Nato s pomočjo zunanjih storitev obiskovalcem pošljemo SMS sporočilo z datumom in časom njihovega srečanja z osebo, ki ima njim najbolj podoben način hoje, hkrati pa vizualiziramo tudi točke korespondence med osebama. Metoda deluje v realnem času, sama ideja prepoznavanja hoje pa ima potencial za uporabo na številnih drugih področjih kot sta zdravstvo in varnost.

• Jezik: Slovenski jezik
• Ključne besede: analiza hoje, interaktivna instalacija, rekonstrukcija, uporabniška izkušnja, novomedijska umetnost, računalniški vid, metoda podpornih vektorjev, nevronske mreže
• Vrsta gradiva: Diplomsko delo/naloga
• Tipologija: 2.11 - Diplomsko delo
• Organizacija: FRI - Fakulteta za računalništvo in informatiko
• Leto izida: 2023
• PID: 20.500.12556/RUL-148943
• COBISS.SI-ID: 164034819
• Datum objave v RUL: 01.09.2023

Sekundarni jezik

• Jezik: Angleški jezik
• Naslov: Reconstruction of an interactive installation with gait recognition

Izvleček

Within the scope of the diploma thesis, the reconstruction of the interactive installation called Matrix of Coincidence from 2006 (author: Sreco Dragan), which is based on the similarity between people's walking patterns, has been implemented. The work is divided into two main parts. The first part presents a method for analyzing walking patterns, their characteristics, and a method for obtaining an identification vector of a person based on these patterns, which works with the help of neural networks. In the second part, we focus on the practical use and implementation of the method used to reconstruct a part of the installation. Within the installation, visitors were able to record their walking patterns and compare them with patterns of other visitors. The similarity search is based on the support vector method, which achieved 81 % accuracy on test data. Then, with the help of external services, we sent visitors an SMS message with the date and time of their meeting with the person who has the most similar walking style, while also visualizing the correspondence points between the individuals. The method works in real-time, and the idea of recognizing walking patterns has potential for use in many other areas such as healthcare and security.

• Ključne besede: gait analysis, interactive installation, reconstruction, user experience, new media art, computer vision, support vector machine, neural networks