Exploring Attentive User Interface Input via Raspberry Pi, based on Face Landmark Detection, Eye Open-Closed Detection and Head Movements Detection

Attentive User Interface (AUI), especially vision-based AUI systems, is a highly developed area of study. Cameras play a crucial role in modeling AUI inputs to monitor and understand user attention during interaction. This enables the optimization of interaction tasks between humans and machines. Typically, AUI systems are designed for desktop computers, employing various detection methods based on computer vision. This research introduces the exploration of AUI input through a low-cost Raspberry Pi 4 embedded system a small-sized system with limited voltage sources. This exploration serves as a reference for AUI design on embedded systems. The proposed AUI input design generating 8 AUI input categories. These include face detection, open or closed eye condition detection, and detection of forward, up, down, left, and right view directions. This study discusses three scenarios of AUI input modeling using various methods to identify the most effective design through Raspberry Pi 4. Based on experimental results and a series of tested tasks, Scenario 3, based on Mediapipe, yielded the best results with an average FPS value of 10 and 100 accuracy for facial landmark detection, detection of open eye conditions, detection of forward, up, down, right, and left gaze directions at various detection reading angles. Copyright: ©2024 The authors. This article is published by IIETA and is licensed under the CC BY 4.0 license.