In certain disabilities, children and young adults can exhibit a range of extreme problem behaviours, including episodes of biting, kicking and self-injurious behaviour which might occur numerous times even in a single day. Existing treatments rely upon a detailed assessment of precisely which behaviour is occurring when, so that the causes of individual patterns of behaviour can be identified and the resulting behaviour addressed.
Presently such assessments are conducted manually over a period of weeks in a clinical setting by specially trained observers. Not only does the intensive resource required mean that this treatment is often unavailable to a large number of children who need it, but the artificial clinical context and the challenges of counting behaviour in real time limits the effectiveness of the treatment itself.
This project is designed with the intention of addressing this state of affairs, by developing a Human Activity Recognition System that uses appropriate sensing technologies in order to track this behaviour. To do so, new methods of privacy sensitive data collection will be developed and existing time-series methods will be adapted to work effectively in this problem space. Data collected through our collaborators, including Georgia Tech in the US, will be used as a basis for developing the system.