Non-stationary BERT: Exploring Augmented IMU Data For Robust Human Activity Recognition

TL;DR

The paper tackles robust, user-specific HAR on mobile devices by introducing OPPOHAR and a lightweight Non-stationary BERT (NS-BERT) with a two-stage training regime. A cross-signal data augmentation based on Factorization Machine design, $fm_{i,j,k} = acc_{i,j} \times gyro_{i,k}$, enhances inter-axis relationships and improves generalization. The NS-BERT uses Series Stationarization and De-stationary Attention to manage IMU non-stationarity, with a Decoder trained by $\mathcal{L}_{MSE}$ to reconstruct sequences, enabling effective pretraining and privacy-preserving on-device finetuning. Experimental results show state-of-the-art performance on OPPOHAR and public HAR datasets, particularly with augmentation and user-specific fine-tuning, underscoring practical impact for energy-efficient, on-device HAR systems.

Abstract

Human Activity Recognition (HAR) has gained great attention from researchers due to the popularity of mobile devices and the need to observe users' daily activity data for better human-computer interaction. In this work, we collect a human activity recognition dataset called OPPOHAR consisting of phone IMU data. To facilitate the employment of HAR system in mobile phone and to achieve user-specific activity recognition, we propose a novel light-weight network called Non-stationary BERT with a two-stage training method. We also propose a simple yet effective data augmentation method to explore the deeper relationship between the accelerator and gyroscope data from the IMU. The network achieves the state-of-the-art performance testing on various activity recognition datasets and the data augmentation method demonstrates its wide applicability.

Figures (3)

• Figure 1: The Overview of our Non-stationary BERT network
• Figure 2: Non-stationary BERT pretraining workflow
• Figure 3: The training loss with epochs.