Human head tracking and gesture recognition are both known problems with solutions using RGB-cameras or an infrared emitter/receiver setup. In this thesis, we propose a method for head tracking and gesture detection using an 8-by-8 infrared sensor array. For this, a novel time-of-flight infrared sensor array is employed, which is both financially and computationally in expensive, while also alleviating privacy concerns due to the very low resolution of the array. The method is split into two parts: first, a human head is detected using circle detection on the filtered combination of depth and amplitude images. If no circle is detected, shape information is used to estimate the position of the head. To reduce false detection and outliers, the movement of the head is tracked over time. Using the depth value of the detected centroid, gesture detection then looks for movement in the given space between the sensor and detected centroid depth (gesture space) and tracks it over five frames. If the major movement direction exceeds a speed of four pixels per second, a gesture is detected. The experiments are split up into field (sensor on a desk in a living room with a window behind the person, daylight) and laboratory (sensor on a turntable in a photography light tent in a lab with artificial ceiling lighting).


Ismail, O. (2024). 3D Head Tracking and Gesture Recognition using an 8-by-8 Array of Infrared Sensors [Diploma Thesis, Technische Universit├Ąt Wien]. reposiTUm. https://doi.org/10.34726/hss.2024.119367