| Summary: | This research paper explores the development and evaluation of a real-time head pose estimation algorithm for visual attention application, addressing challenges in varying conditions, complex poses, and partial occlusion. Leveraging the OpenCV library and Haar cascade classifier, the algorithm was implemented and tested with a smartphone camera setup. The study involved 10 subjects under different conditions, revealing high accuracy in controlled scenarios. The methodology incorporated innovative features, including cascade classifiers for diverse facial orientations. Results indicated varying accuracy influenced by environmental factors and subject movements. The average accuracy of the developed algorithm applied to various testing conditions is more than 84% for head pose estimation and more than 90% for visual attention. The findings contribute insights into algorithm efficacy, showcasing potential applications in fields of healthcare, therapy, driving monitoring and others. Overall, this research lays a foundation for robust head pose estimation systems with real-world adaptability. © 2024 Seventh Sense Research Group®.
|
|---|