MECKD: Deep Learning-Based Fall Detection in Multilayer Mobile Edge Computing With Knowledge Distillation

TL;DR

The paper tackles real-time fall detection in aging populations by balancing accuracy and latency in a wearable sensing setting. It proposes a multilayer mobile edge computing (MLMEC) framework that distributes DL models across edge devices, MEC servers, and the cloud, using knowledge distillation to transfer learning from powerful back-end models to lightweight front-end models. A threshold-based upward judgment mechanism and staged KD (including teacher-TA-student paths) improve front-end detection and reduce data transmissions, achieving notable accuracy gains on SisFall and FallAllD and substantial latency reductions. Experiments on public datasets demonstrate that the triple-layer MLMEC with KD, particularly with ResNet18 as teacher, MobileNetV3 as TA, and CNN as student, delivers the best trade-off between accuracy and efficiency. The approach offers a practical path toward real-time, privacy-conscious FD in resource-constrained mobile environments, with implications for deployment and future extensions to other biosignals.

Abstract

The rising aging population has increased the importance of fall detection (FD) systems as an assistive technology, where deep learning techniques are widely applied to enhance accuracy. FD systems typically use edge devices (EDs) worn by individuals to collect real-time data, which are transmitted to a cloud center (CC) or processed locally. However, this architecture faces challenges such as a limited ED model size and data transmission latency to the CC. Mobile edge computing (MEC), which allows computations at MEC servers deployed between EDs and CC, has been explored to address these challenges. We propose a multilayer MEC (MLMEC) framework to balance accuracy and latency. The MLMEC splits the architecture into stations, each with a neural network model. If front-end equipment cannot detect falls reliably, data are transmitted to a station with more robust back-end computing. The knowledge distillation (KD) approach was employed to improve front-end detection accuracy by allowing high-power back-end stations to provide additional learning experiences, enhancing precision while reducing latency and processing loads. Simulation results demonstrate that the KD approach improved accuracy by 11.65% on the SisFall dataset and 2.78% on the FallAllD dataset. The MLMEC with KD also reduced the data latency rate by 54.15% on the FallAllD dataset and 46.67% on the SisFall dataset compared to the MLMEC without KD. In summary, the MLMEC FD system exhibits improved accuracy and reduced latency.

Figures (10)

• Figure 1: Typical two-tier and the proposed MLMEC FD frameworks.
• Figure 2: The application scenario of the proposed MLMEC FD system.
• Figure 3: The flow chart of the proposed MLMEC FD system.
• Figure 4: Impact-defined window on a fall trial re19_An_analysisReview1_Add15, where the X-axis represents the sampling time (unit: second), and the Y-axis represents the acceleration momentum (unit: gram).
• Figure 5: The illustration of the $Norm_{xyz}$ and $Norm_{hori}$ distribution of falls and ADLs re18_The_Methodsre19_An_analysis (a) The boxplots of maximum and minimum values of $Norm_{xyz}$ for falls and ADLs; (b) The boxplots of maximum and minimum values of $Norm_{hori}$.