Action quality assessment methods for ASD behaviour evaluation

Dinghuang Zhang, Dalin Zhou, Honghai Liu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

26 Downloads (Pure)

Abstract

Given the current increasing prevalence of autism, expensive and time-consuming manual diagnosis is highly detrimental to the management of the condition. With the development of computer-based methods of human behavioural analysis, these methods are expected to provide more accurate, objective and reproducible methods of early screening and diagnosis of autism. To advance the field of behavioural quantification in autism research, this study utilises human skeletal behavioural data from publicly available autism datasets and ADOS scores from clinical professionals in a first attempt to build deep neural networks that can predict ADOS scores from behavioural data using the AQA approach. This paper finds a moderately correlated between the ground truth ADOS score and the predicted ADOS score, it reveals the potential use of the AQA method in ASD diagnoses.
Original languageEnglish
Title of host publication2023 International Conference on Machine Learning and Cybernetics (ICMLC)
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages6
ISBN (Electronic)9798350303780
ISBN (Print)9798350303797
DOIs
Publication statusPublished - 28 Nov 2023
EventThe 22nd International Conference on Machine Learning and Cybernetics - Adelaide, Australia
Duration: 9 Jul 202311 Jul 2023

Publication series

Name2023 International Conference on Machine Learning and Cybernetics (ICMLC)
PublisherIEEE
ISSN (Print)2160-133X
ISSN (Electronic)2160-1348

Conference

ConferenceThe 22nd International Conference on Machine Learning and Cybernetics
Abbreviated titleICMLC
Country/TerritoryAustralia
CityAdelaide
Period9/07/2311/07/23

Keywords

  • AQA
  • ASD
  • behaviour evaluation
  • CNN-LSTM
  • skeleton

Cite this