A Novel Machine Learning Based Wearable Belt For Fall Detection
Kimaya Desai, Pritam Mane, Manish Dsilva, Amogh Zare, Parth Shingala, Dayanand Ambawade
Abstract
Falls are a major cause of hip fractures in elderly people. Such fractures take a lot of time to heal even after a surgery is performed. In addition to this, the majority of such injuries prove to be fatal due to the lack of quick communication and immediate medical response. This situation is pretty common in today's world where the elderly are most of the time unattended at home. Owing to this need, we have designed a system to detect such falls leveraging machine learning and signal processing algorithms deployed over a simple 32-bit micro-controller. To achieve higher accuracy, we prepared our custom dataset of various types of fall as well as other daily routine activities. Our device informs the close relatives/family via a GSM Module when a fall is detected. The main purpose of our system is to detect a fall and trigger the alert system and take immediate action to minimize the impact of the fall. Our system has been able to detect a fall within 0.25 seconds with high accuracy. This can further be used to develop a real-time safety mechanism gear to minimize the injury in case of a fall.