Emerging Trends in Machine Learning-Driven Fall Detection: A Review of Deep Learning Architectures, Sensor Fusion and Edge Computing for Real-Time Applications

The occurrence of falls creates danger for people who are older adults and those with limited mobility because it produces serious medical outcomes such as hospitalization with extended disabilities. The existing methods for fall detection, which include threshold-based systems and human observation, fail to deliver satisfactory results because they trigger many incorrect alarms and lack personalization capabilities. Current advances in real-time fall detection as well as prevention systems emerged through the combination of machine learning techniques with Internet of Things devices. The paper combines analysis from ten essential studies published between 2018 to 2025 which study deep learning network progress through Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) and Transformer-based algorithms. The paper investigates how wearable and environmental sensors together with multimodal sensor fusion and edge computing enhance both detection accuracy along with operational effectiveness. An evaluation of present techniques demonstrates their operational potential together with defined weaknesses and implementation capabilities in real-world scenarios. This review examines crucial issues surrounding privacy in addition to identifying performance and adaptability challenges that affect different population groups. Future studies should create individualized models and should focus on real-time edge processing and establish more secure protection systems for data security.