Elderly Fall Detection Method Using Threshold Based Method and Transfer Learning
Yuzhi Ma, Lei Tang, Jiahui Liu · Frontiers in artificial intelligence and applications · 2024
The lack of elderly data is one of the most challenging issues in elderly fall detection. And the simulated young data is usually used in fall detection researches. In this paper, a wearable elderly fall detection method based on threshold based method (TBM) and transfer learning is proposed to shorten the distribution difference between young data and elderly data. In the process of training, labeled young data (source domain) and unlabeled elderly data (target domain) are utilized so as to weak the adverse effects of distribution difference between young data and elderly data. Finally, the better performance with sensitivity 98.63% and specificity 98.00% is illustrated with elderly data in Sisfall dataset.