Mobile Sensor-Based Fall Detection Framework
Disciplines
Engineering | Health Information Technology
Abstract (300 words maximum)
Fall is a major concern among elderly population. Accidental fall if unattended for a long time, may lead to severe injuries and disability. Prompt detection of fall is an important research problem, particularly in the homecare settings for elderly citizens, where not enough service providers are available to monitor their health and welfare daily. There are some available fall detection approaches, however, they are either expensive or the accuracy of fall detection is not satisfactory. In this paper, we develop a low-cost fall detection framework using Android phones built-in sensors with the goal of detecting falls and notifying to emergency responders. We generate a dataset and train 3 popular machine learning algorithms to detect fall events: Logistic Regression, Naïve Bayes and Neural Network. Our study shows the performance comparison of the learning algorithms. The evaluation results show that our approach can successfully detect fall and neural network-based technique can perform better than other learning techniques.
Academic department under which the project should be listed
CCSE - Information Technology
Mobile Sensor-Based Fall Detection Framework
Fall is a major concern among elderly population. Accidental fall if unattended for a long time, may lead to severe injuries and disability. Prompt detection of fall is an important research problem, particularly in the homecare settings for elderly citizens, where not enough service providers are available to monitor their health and welfare daily. There are some available fall detection approaches, however, they are either expensive or the accuracy of fall detection is not satisfactory. In this paper, we develop a low-cost fall detection framework using Android phones built-in sensors with the goal of detecting falls and notifying to emergency responders. We generate a dataset and train 3 popular machine learning algorithms to detect fall events: Logistic Regression, Naïve Bayes and Neural Network. Our study shows the performance comparison of the learning algorithms. The evaluation results show that our approach can successfully detect fall and neural network-based technique can perform better than other learning techniques.