Hybrid PSO-InceptionNet for Wearable Sensor Data Analysis in Elderly Fall Risk Assessment
B Sabeena, Ranjeeth Kumar C, Kandala Vaishnavi, Satya Subrahmanya Sai Ram Gopal Peri, Shamna Parveen S, N. V. Keerthana · 2025
Fall prediction in the elderly is a key aspect of a proactive healthcare system, as such accidents most often result in critical injuries, and the quality of life of the individual is greatly affected. This study proposes a hybrid arrangement, that is to say, combining the Particle Swarm Optimization (PSO) with InceptionNet, and through this, enhancing the accuracy and efficiency of the assessment of fall risk using wearable sensor data. The PSO algorithm optimizes parameters and selects the best features, making sure that only the most relevant sensor data is tested, while the learning architecture of InceptionNet extracts the hierarchical spatial-temporal patterns from the motion signals. The new model is programmed to be capable of refacing real-time sensor data, and it provides a learnable framework to improve the generalization across different subjects. The experimental results show that the Hybrid PSO-InceptionNet model outperforms the traditional supervisor learning methods based on precision, sensitivity, and computational efficiency. The study's contributions are the use of PSO for feature optimization, the new application of InceptionNet in wearable sensor analysis, and the enhancement of the fall risk prediction framework with high robustness and adaptability. This work has the potential to support early intervention strategies and personalized fall prevention systems for the elderly.