Persons with cognitive disabilities such as Autistic Spectrum Disorders (ASD) and intellectual disabilities tend to have problems in sequencing and coordinating steps in the execution of basic Activities of Daily Living (ADLs) due to limited capabilities in cognitive functioning. In order to successfully perform basic ADLs, these persons are highly reliant on the assistance of a human caregiver. This leads to a decrease or even a loss of independence for care recipients and imposes a high burden on caregivers.
Assistive Technology for Cognition (ATC) aims to compensate for decreased cognitive functions. ATC systems provide automatic assistance in the execution of ADLs by delivering appropriate prompts which enable the user to perform basic ADLs without any assistance of a human caregiver. This leads to an increase of the user's independence and to a relief of caregiver's burden.
In this thesis, we describe the design, development and evaluation of a novel ATC system. The TEBRA (TEeth BRushing Assistance) system supports persons with moderate cognitive disabilities in the execution of brushing teeth by providing audio-visual prompts to the user.
In order to reveal the characteristics of the task and the involved users, we conduct Interaction Unit (IU) analysis, a structured method of task analysis. We iteratively refine the initial design decisions based on the results of IU analysis in intermediate evaluations where we follow a user-centered design: in a Wizard of Oz study, we evaluate the reaction behaviors of persons with cognitive disabilities to system prompts. In an interview study, we ask professional caregivers about appropriate modalities and content of prompts.
We incorporate the design decisions into the implementation of the TEBRA system. A main requirement for the acceptance of an ATC system is context awareness: an explicit feedback from the user is not necessary in order to provide appropriate assistance. We allow for context awareness by implementing a user behavior recognition component which deals with the variations in the execution of behaviors such as different movement characteristics and different velocities: we infer user behaviors based on states of objects involved in the task which we apply in a Bayesian Network classification scheme. A dynamic timing model allows for different velocities of users and adapts to a user's velocity during a trial.
We evaluate a fully functioning prototype of the TEBRA system in a study with persons with cognitive disabilities. The main aim of the study is to analyze the technical performance of the TEBRA system and the user's behavior in the interaction with the system with regard to the main hypothesis: Is the TEBRA system able to increase the independence of users in the execution of brushing teeth?