Enhancing Fall Detection in Assisted Living Using Deep Learning Approach
S Sangeetha, K. Anitha, N. Chidambararaj, Swagata B. Sarkar, Karthika, S Srimathi · 2025
Aged and disabled people living alone are at a greater risk of falls. Older folks are more likely to fall and get injured, because they are weaker, unstable, and slower to react. In reality, elderly folks worry more about fall effects than prevalence. The WHO says, falls are the greatest cause of injury and death of old people worldwide. Globally, the senior population is growing. Thus, fall detection is becoming an assistive living tool for those people. Computer vision and deep learning are used extensively in assisted living. Most research and companies offer a variety of workable methods to safeguard the elderly and those caring for them, against falls. Falls are recognized, and alerting systems are activated to call for aid. A fall detection algorithm examines human body shape in video frames. Posture estimation, the angle and separation between the vector created by the head-centroid of the identified facial image and the middle hip of the human body, and the vector aligned with the centre hip's vertical axis are used to generate novel, distinguishable image characteristics. Using estimated angle and distance sequences, a Bi-directional Long-Short-Term Memory (Bi-LSTM) network with attention mechanism is trained to distinguish fall-related activities from non-fall-related activities. This research employs a deep neural network and a wrist-worn gadget to recognize abnormal walking patterns. Convolution and Bi-LSTM layers facilitate temporal characteristic learning from multiple detector inputs.