Alzh-care: - Knowledge - driven intelligent system to assist Alzheimer Patients

• Published in: 2019 1st International Conference on Advanced Technologies in Intelligent Control, Environment, Computing & Communication Engineering (ICATIECE) pp. 114 - 120
• Main Authors: Rashmi, S.R., Suha, Sameia, Krishnan, R., Ramesh Babu, D.R.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.03.2019
• Subjects: Activity recognition; Domain Ontology; Hidden Markov models; Inference; Intention Recognition; Monitoring; Ontologies; Senior citizens; Smart homes; State Machine; Task; Task analysis
• DOI: 10.1109/ICATIECE45860.2019.9063626

Alzh-care is an Alzheimer patients assistive system. This system assists the patients by recognizing their intention on the activity and guides them to the completion of that task. It helps the patients in completing their day-today tasks by recognizing their intent of the task at the early stage. In this work, Intention recognition is achieved through the sensor based activity recognition process. These kinds of systems have far reaching applications in the domains like assisted living; healthcare monitoring and they offer lots of open challenges to solve in the arena of research and development. This paper discusses the implementation of the Alzh-care assistive system using sensors for the day-to-day activities like eating meals and taking medicines etc., We have used knowledge driven approach for representing the domain knowledge on the day-to-day activities is as ontology. This in turn is referred to know the intentions of the patients and voice assist them in completing the intended task. We have tried state machine approach to infer and assist on the task. Knowledge driven approach has proven to perform better than the data driven approaches as the latter suffers from the cold start and reliability issues.We have tabulated the trials and noted the success rate of recognition of the intent for each of the activity experimented. This system is proven to be 85% to 90% accurate for predicting the intentions in the finite disambiguous tasks, owing the other 5% to 10% of the failure caused by the sensors used.