OAK-D as a Platform for Human Movement Analysis: A Case Study

• Daniel Perazzo Universidade Federal de Pernambuco - UFPE
• Natalia Souza Soares Universidade Federal de Pernambuco - UFPE
• Victor Gouveia de Menezes Lyra Universidade Federal de Pernambuco - UFPE
• Gustavo Camargo Rocha Lima Universidade Federal de Pernambuco - UFPE
• Alana Elza Fontes da Gama Universidade Federal de Pernambuco - UFPE
• Joao Marcelo Xavier Natario Teixeira Universidade Federal de Pernambuco - UFPE
• Veronica Teichrieb Universidade Federal de Pernambuco - UFPE

Resumo

3D interaction is an essential manner of interaction in VR environments. Many devices enable analysis and use of human inputs in these settings, such as the Kinect™ from Microsoft. One of the latest additions is the OpenCV AI Kit with Depth or OAK-D. This paper analyzes the OAK-D and proposes a case study based on a Motion Analysis framework created for at-home workouts to explore its capabilities in a 3D interaction setting. Our results show that the OAK-D is well suited for this kind of task, achieving 7 FPS with a Raspberry Pi 4 and 15 FPS with a laptop. Furthermore, it could detect and correct wrong movements while the user was doing it and count repetitions for the squat and side lateral raise.

Palavras-chave: Human-computer interaction, Spatial AI, Motion Analysis, 3D Interaction, Embedded AI, Pose Estimation