Research on Age-Friendly Home Design Based on Regularized Non-negative Matrix Factorization Algorithm

Population aging has become a major issue in Asian countries. Advances in computer algorithms have enabled the design and development of smart home and elderly care devices. Given the limitations of existing fall detection algorithms, this paper proposes a method for an intelligent fall prevention system for the elderly, focusing on non-negative matrix factorization visual correction. This research addresses the challenges of automatically adjusting the parameters of visual detection systems in real-world scenarios. The method combines LSTM with non-negative matrix factorization visual optimization algorithms to eliminate environmental noise's impact on visual analysis. It iteratively estimates background noise power levels based on algorithm reference layers, achieving accurate detection of elderly fall behaviors and posture features. The proposed method compensates for inaccuracies in posture description and significant tracking errors of traditional integer-order admittance control. Experimental results show that, compared to other methods, the proposed method offers superior fall posture detection performance. Therefore, it can be effectively applied to elderly care smart home systems, providing new strategies for optimizing smart furniture performance.