Vision-based Action Understanding for Assistive Healthcare: A Short Review

Md Atiqur Rahman Ahad, Anindya Das Antar, Omar Shahid; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 1-11

The scarcity of trained therapist, economic imbalance, and an increasing amount of elderly people are the reasons for poor rehabilitation treatment and inadequate healthcare facilities in many countries. Vision-based rehabilitation treatment, monitoring daily living, and advanced healthcare can improve technology that allows people with an injury to practice intense movement training without taking help from a therapist daily. This technology has remarkable basic notable benefits as vision-based systems are non-contact, precise, immune to electromagnetic interference, nondestructive, and they can be used for long range and multiple target monitoring. The objective of this survey paper is devoted to exhibiting a summary of the challenges and difficulties in this domain along with some solutions. Besides, in order to guide the researchers in this field, we have discussed available sensing devices in the field of computer vision that can be used for taking data in hospitals and rehabilitation centers. We have also analyzed some benchmark datasets regarding gestures, medical activities, sports and exercise actions, 3D actions, and so on with relevant information. Moreover, this article also provides a comparison among existing research works on some benchmark datasets related to this field of research.