A portable upper extremity rehabilitation device

We present artificial intelligence techniques employed in the development of an effective rehabilitation device, a "smart can", that assesses hand function for individuals with stroke, spinal cord injury (SCI), traumatic brain injury (TBI), or other neurologic injuries. Currently, at the end of their formal therapies, many individuals with these pathologies are typically provided only with written home exercise programs prescribed by their therapists. It is not clear how performance changes are measured over time at home and it is therefore not surprising that these written exercises are commonly discontinued shortly after the formal therapies end. To this end, using Al-based methods, we have designed a rehabilitation device that specifically explores the use of a new type of performance measurement, "the jerk score", so that the device can be eventually deployed in a home setting providing the user with a direct measure of the quality of the movements made by the affected hand. Therefore, using computer vision based techniques such as object detection, incremental visual tracking, activity recognition, and 3D virtual augmentation, we successfully demonstrate the efficacies of the system in the lab setting.