Hand gesture detection in the hand movement test for the early diagnosis of dementia
Year: 2021
Type: preprint
Abstract: Collecting hands data is important for many cognitive studies, especially forsenior participants who has no IT background. For example, alternating hand
movements and imitation of gestures are formal cognitive assessment in the
early detection of dementia. During data collection process, one of the key
steps is to detect whether the participants is following the instruction
correctly to do the correct gestures. Meanwhile, re-searchers found a lot of
problems in TAS Test hand movement data collection process, where is
challenging to detect similar gestures and guarantee the quality of the
collect-ed images. We have implemented a hand gesture detector to detect the
gestures per-formed in the hand movement tests, which enables us to monitor if
the participants are following the instructions correctly. In this research, we
have processed 20,000 images collected from TAS Test and labelled 6,450 images
to detect different hand poses in the hand movement tests. This paper has the
following three contributions. Firstly, we compared the performance of
different network structures for hand poses detection. Secondly, we introduced
a transformer block in the state of art network and increased the
classification performance of the similar gestures. Thirdly, we have created
two datasets and included 20 percent of blurred images in the dataset to
investigate how different network structures were impacted by noisy data, then
we proposed a novel net-work to increase the detection accuracy to mediate the
influence of the noisy data.
Source: arXiv (Cornell University)
Authors: Guan Huang, Son N. Tran, Quan Bai, Jane Alty
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W3209693631