Stay Informed:

COVID-19 (coronavirus) information
Zoom Links: Zoom Help | Teaching with Zoom | Zoom Quick Guide

Advancement: Vision-based Advanced Head Based Pointing

Speaker Name: 
Muratcan Cicek
Speaker Title: 
PhD Student
Speaker Organization: 
Computer Engineering PhD
Start Time: 
Tuesday, June 2, 2020 - 3:00pm
End Time: 
Tuesday, June 2, 2020 - 4:00pm
Location: 
Zoom - https://ucsc.zoom.us/j/98016481941

Abstract: Head-based pointing (HBP) can be an effective alternative to classical computer input methods such as mouse or trackpad. In particular, for individuals who cannot use the mouse, trackpad, or keyboard due to physical disabilities, and who cannot use speech-based input due to dysarthria or aphasia, HBP represents one of the few computer interface techniques available. Recent advances in computer vision technology have enabled vision-based HBP interaction on commodity laptop computers and mobile devices using built-in hardware. Current strategies for mapping visual data to pointer control are based on simple models inherited from traditional mouse or tracking pad interface mechanisms. I claim that these simple models are not sufficient for natural head-based pointing interactions and that more complex mapping functions are needed. I propose to develop and test a system that uses a neural network to directly map visual input to the pointer location. This network will be trained on data collected in a carefully designed user study, which will also involve individuals with motor impairment. In this presentation, I will present prior work on HBP that I completed during an internship at several companies, describe the proposed data acquisition study (including results from a pilot study), and outline the architecture of the proposed HBP system, along with the proposed user study to validate the system.

Event Type: 
Adancement/Defense
Advisor: 
Roberto Manduchi
Graduate Program: 
Computer Engineering PhD