Banca de DEFESA: WINNIE DE LIMA TORRES

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : WINNIE DE LIMA TORRES
DATE: 07/07/2026
TIME: 14:00
LOCAL: Remoto (meet)
TITLE:

Distributed Framework for Gesture Interaction in Augmented Reality Applied to Physical Device Control

KEY WORDS:

Augmented reality; Gestural interaction; Distributed architecture; Gesture-based control

PAGES: 210
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:

Controlling physical devices through gestures in augmented reality requires more than
recognizing the hand configuration in an image. In interactive applications, especially
when a physical system is actuated, small variations in landmarks, temporary tracking
losses, lighting changes, occlusions, and individual differences in gesture execution may
lead to unstable or unintended commands. This thesis addresses this problem and propo-
ses a distributed architecture for gestural interaction in augmented reality applied to the
control of physical devices. The proposed architecture organizes the system into layers
responsible for perceptual acquisition, geometric-temporal processing of hand landmarks,
interaction interpretation, distributed communication, and physical actuation. The pro-
cessing model uses spatial normalization, geometric descriptors, and temporal filtering to
reduce oscillations in perceptual data. The interaction model structures execution through
operational states, incorporating temporal persistence, confirmation, cancellation, rever-
sal, and continuous parameter adjustment. The architecture was implemented in an expe-
rimental WebAR-based platform using an augmented reality device, MediaPipe Hands for
landmark extraction, a communication middleware, and a physical lighting panel driven
by an embedded device. The experimental evaluation was conducted in six test scena-
rios involving individual selection, collective selection, entry into editing mode, intensity
adjustment, confirmation, and cancellation. The results indicated an overall success rate
of 93.33%, a false activation rate of 1.67%, a mean global intensity error of 1.68%, and
a mean total latency of 143.47 ms between validated command and physical actuation.
Temporal filtering reduced the variability of the signal associated with the pinch gesture
by approximately 50%. These results show that the integration of geometric-temporal
processing, state-based interaction, and distributed communication contributes to making
gesture-based control in augmented reality more stable, predictable, and suitable for ex-
perimental applications involving physical devices.

COMMITTEE MEMBERS:
Presidente - 2579664 - ALLAN DE MEDEIROS MARTINS
Interno - 347628 - ADRIAO DUARTE DORIA NETO
Externo à Instituição - GLAUBER RODRIGUES LEITE - UFAL
Externo à Instituição - TÁCITO TRINDADE DE ARAÚJO TIBURTINO NEVES - UFPB
Externo à Instituição - ÍCARO BEZERRA QUEIROZ DE ARAÚJO - UFAL