An Artificially Intelligent Wearable Device for Dementia Patients
Arshad Mohammed, Gazanfur Ali Mohammed · 2019
The rate at which people contract dementia is increasing rapidly, with it afflicting 75 million patients by the year 2030. We need to develop the technology to improve the quality of life for these patients while we search for a cure. One of the most frustrating symptoms is constantly losing items in unusual places. Dementia patients are likely to spend hours searching for items daily. However, the only solution in the market is mainly ineffective, GPS trackers. They can track individual persons or items; however, they are small enough to get misplaced and can only track an object at a time. The user also has to remember how to use the device and be aware of their memory loss. This is especially difficult for patients with executive cognitive function (ECF) deficits. We propose a unique wearable device that uses object detection to track pretrained items and learn to detect new and familiar objects. We demonstrate that by linking together an SSD(Single Shot Detection) neural network with saliency detection and k-means clustering algorithm and internal positioning system, we were able to build such a device. Most importantly, the device has high precision and recall on trained items and can sufficiently distinguish between highly common objects and miscellaneous items and draw bounding boxes around the valuable data to train with later on.