PreFallKD: Pre-Impact Fall Detection via CNN-ViT Knowledge Distillation

TL;DR

PreFallKD introduces a knowledge distillation approach combining CNN and ViT models to improve pre-impact fall detection accuracy while maintaining low computational complexity for wearable devices.

Contribution

It presents a novel CNN-ViT knowledge distillation method for efficient pre-impact fall detection, balancing performance and resource constraints.

Findings

Achieves 92.66% F1-score on KFall dataset

Provides 551.3 ms lead time for fall warning

Outperforms existing state-of-the-art models

Abstract

Fall accidents are critical issues in an aging and aged society. Recently, many researchers developed pre-impact fall detection systems using deep learning to support wearable-based fall protection systems for preventing severe injuries. However, most works only employed simple neural network models instead of complex models considering the usability in resource-constrained mobile devices and strict latency requirements. In this work, we propose a novel pre-impact fall detection via CNN-ViT knowledge distillation, namely PreFallKD, to strike a balance between detection performance and computational complexity. The proposed PreFallKD transfers the detection knowledge from the pre-trained teacher model (vision transformer) to the student model (lightweight convolutional neural networks). Additionally, we apply data augmentation techniques to tackle issues of data imbalance. We conduct the experiment on the KFall public dataset and compare PreFallKD with other state-of-the-art models. The experiment results show that PreFallKD could boost the student model during the testing phase and achieves reliable F1-score (92.66%) and lead time (551.3 ms).