Motion Sensor–Based Fall Prevention for Senior Care: A Hidden Markov Model with Generative Adversarial Network Approach.

This article focuses on the development and evaluation of a novel fall prevention framework for chronic condition–related mobility threats using wearable motion sensor data. The framework integrates a hidden Markov model with a generative adversarial network (HMM-GAN) to automatically extract temporal and sequential patterns from sensor signals without manual feature engineering, and employs logistic regression to decide when to trigger protective devices with sufficient lead time to prevent falls. Evaluated on two large-scale, labeled data sets in both semisupervised and supervised modes, HMM-GAN outperformed state-of-the-art sensor analytics and fall prevention models in accurately recognizing snippet states and successfully preventing falls. The study contributes methodological advances to information systems (IS) research by introducing a new expectation-maximization algorithm for joint HMM-GAN parameter learning and demonstrates practical benefits for senior care, including potential economic savings and improved safety.