Diffusion-Model-Based Contrastive Learning for Human Activity Recognition

• Published in: IEEE internet of things journal 2024-10, Vol.11 (20), p.33525-33536
• Main Authors: Xiao, Chunjing, Han, Yanhui, Yang, Wei, Hou, Yane, Shi, Fangzhan, Chetty, Kevin
• Format: Article
• Language: English
• Subjects: Activity recognition; Adaptation models; Adaptive algorithms; Adaptive sampling; Computational modeling; Contrastive learning; Data augmentation; Data models; Diffusion models; diffusion probabilistic models; Fluctuations; Human activity recognition; Human motion; Machine learning; self-supervised learning; Training; Training data; WiFi channel state information (CSI)
• ISSN: 2327-4662; 2327-4662
• DOI: 10.1109/JIOT.2024.3429245

WiFi channel state information (CSI)-based activity recognition has sparked numerous studies due to its widespread availability and privacy protection. However, when applied in practical applications, general CSI-based recognition models may face challenges related to the limited generalization capability, since individuals with different behavior habits will cause various fluctuations in the CSI data and it is difficult to gather enough training data to cover all kinds of motion habits. To tackle this problem, we design a diffusion model-based contrastive learning framework for human activity recognition (CLAR) using WiFi CSI. On the basis of the contrastive learning framework, we primarily introduce two components for CLAR to enhance the CSI-based activity recognition. To generate diverse augmented data and complement limited training data, we propose a diffusion model-based time series-specific augmentation model. In contrast to typical diffusion models that directly apply conditions to the generative process, potentially resulting in distorted CSI data, our tailored model dissects these condition into the high-frequency and low-frequency components, and then applies these conditions to the generative process with varying weights. This can alleviate the data distortion and yield high-quality augmented data. To efficiently capture the difference of the sample importance, we present an adaptive weight algorithm. Different from the typical contrastive learning methods which equally consider all the training samples, this algorithm adaptively adjusts the weights of positive sample pairs for learning better data representations. The experiments suggest that the CLAR achieves significant gains compared to the state-of-the-art methods.