Vision-augmented home-based stroke rehabilitation device
MetadataShow full item record
The purpose of this paper is to highlight the key role that computer vision and graphics play in the implementation and deployment of a novel and innovative assistive technology, aimed at significantly improving compliance for home-based, post-stroke rehabilitation. In 2010, approximately 17 million people had a stroke worldwide and over 33 million people had previously had a stroke and were still alive. In 2013, stroke was the second most frequent cause of death after coronary artery disease, accounting for 6.4 million deaths globally. Lasting disability affects 75% of stroke survivors enough to decrease their employability. Stroke rehabilitation services are thus aimed at reducing impairment after a stroke and improving functional mobility. Here, we present computer vision techniques employed in the development of a novel and innovative, home-based stroke rehabilitation assistive device, a "smart can". Currently, at the end of their formal therapies, individuals with stroke are typically provided only with written home exercise programs prescribed by their therapists, and these are commonly discontinued. In general, compliance with written home exercise is low, but compliance with home exercise tools has been shown to be high. To this end, an integrated team of scientists has designed and deployed an assistive technology device, aimed at improving compliance primarily via objective feedback and personalization. This research work therefore presents how techniques such as object detection, incremental visual tracking, activity recognition, and 3D virtual augmentation are exploited in the context of enhancing objective feedback on exercise performance and tailoring exercise programs to appropriately challenge participants. We successfully demonstrate the efficacies of the components of the system in the lab setting and going forward, additional usability tests will be performed to optimize the system to the specific needs of its targeted users.