Interface for efficient communication utilizing user gaze tracking

Komunikacija z računalniškimi sistemi je primarno načrtovana za vhodne naprave, ki zahtevajo fizično upravljanje. Te niso primerne v vseh okoliščinah. Zlasti motorično omejeni uporabniki potrebujejo alternativne interakcijske metode. V magistrski nalogi se osredotočimo na interakcije, ki izhajajo izključno iz uporabnikovega pogleda. Lastni pogled pogosto lahko nadzorujemo, tudi ko sistema ne zmoremo upravljati z običajnimi pristopi. V okviru dela raziščemo najbolj primerne interakcijske metode in razvijemo svojo metodo dvonivojskega križanja. Ta ne sloni na standardnem principu aktivacijskega časa. Implementiramo metodo izbire z aktivacijskim časom, dvonivojsko drevesno izbiro, prilagajanje aktivacijskega časa in dvonivojsko križanje. Razvijemo aplikacijo za pisanje besedila z implementiranimi metodami. Integriramo in testiramo tri sledilne sisteme na osnovi točnosti in natančnosti meritev. Z uporabniškimi testi ocenimo učinkovitost pisanja s posamezno interakcijsko metodo. Z enonivojsko metodo dosežemo mediano 7,02 besed na minuto. Z metodo dvonivojskega križanja dosežemo 3.74 besed na minuto. S subjektivno oceno jo izbere 31% testnih uporabnikov kot najbolj primerno metodo za pisanje.Human-computer interaction systems are primarily designed for devices that require physical intervention. Such devices are inappropriate in certain circumstances. It is important to support alternative solutions, especially for the physically disabled. This work focuses on interactions based solely on the user\u27s gaze. In most cases, we have control over our gaze, even when standard interactions are not applicable. We explore the most appropriate interaction methods and develop our own method called two-level crossing, which is not based on dwell time. We implement methods based on dwell time selection, two-level tree-based selection, dwell time adaptation, and two-level crossing. We develop a writing application that includes the implemented methods. We integrate and test three eye-tracking systems for accuracy and precision. We evaluate all interaction methods with user tests. The single-level dwell time selection method achieves a median of 7.02 words per minute. The two-level crossing achieves a median of 3.74. Based on subjective evaluation, 31% of users select the method as the most appropriate for writing