Research on Cloud-Edge Collaborative Fall Detection Methods for Personalized Applications

To improve the accuracy of fall detection and address the performance degradation of detection algorithms caused by individual behavior differences in practical applications, this paper proposes a cloud-edge collaborative fall detection method tailored for user personalization. First, a lightweight convolutional neural network suitable for resource-constrained edge devices is proposed to capture multi-scale features in multichannel sensor data sequences for fall detection. Then, a cloud-edge collaborative user personalization optimization algorithm is introduced, which employs personalized user data to expand its applicability. The experimental results show that the proposed method can effectively improve the accuracy in practical applications.