Utilizing Machine Learning to Recognize Human Activities for Elderly and Homecare

Dementia is a progressive disorder associated with age, which is characterized by deterioration of individuals' cognitive functions such as the ability to perform routine tasks. With the increase of human life expectancy, the prevalence of dementia patients will reach 152 million in 2050. Unfortunately, there is no treatment available to cure dementia or alter the course of its progression. However, there is an area of support for patients and caregivers to assist daily living. Technological devices and applications are increasingly advancing, exploiting sensory data for dementia patients and homecare using smartphones to permit monitoring of their activities.

This paper uses the labeled dataset besides comparing the 3-classification algorithm to evaluate whether or not the algorithms deployed can classify the activities with high accuracy.

A public data is used to classify human activities into one of the six activities, BigML platform is used to build machine learning models. Results show that machine learning algorithms can achieve high accuracy. The activity recognition algorithms are highly accurate using ridged regression and deep neural networks, with almost all activities being recognized correctly over 98% of the time.

An application of smartphones can be utilized for human activities monitoring by proposing a high level for dementia patients and homecare monitoring services. Using this service, the patients only need to carry the smartphone, and their caregivers simply need to use the application that monitors their patients' activities.