Semantically Encoding Activity Labels for Context-Aware Human Activity Recognition
Wen Ge, Guanyi Mou, Emmanuel Agu, Kyumin Lee · 2025
Prior work has primarily formulated Context-Aware Human Activity Recognition (CA-HAR) as a multi-label classification problem, where model inputs are time-series sensor data and target labels are binary encodings representing whether a given activity or context occurs. These CA-HAR methods either predicted each label independently or manually imposed relationships using graphs. However, both strategies often neglect an essential aspect: activity labels have rich semantic relationships. For instance, walking, jogging, and running activities share similar movement patterns but differ in pace and intensity, indicating that they are semantically related. Consequently, prior CA-HAR methods often struggled to accurately capture these inherent and nuanced relationships, particularly on datasets with noisy labels typically used for CA-HAR or situations where the ideal sensor type is unavailable (e.g., recognizing speech without audio sensors). To address this limitation, we propose Semantically Encoding Activity Labels (SEAL), which leverage Language Models (LMs) to encode CA-HAR activity labels to capture semantic relationships. LMs generate vector embeddings that preserve rich semantic information from natural language. Our SEAL approach encodes input-time series sensor data from smart devices and their associated activity and context labels (text) as vector embeddings. During training, SEAL aligns the sensor data representations with their corresponding activity/context label embeddings in a shared embedding space. At inference time, SEAL performs a similarity search, returning the CA-HAR label with the embedding representation closest to the input data. Although LMs have been widely explored in other domains, surprisingly, their potential in CA-HAR has been underexplored, making our approach a novel contribution to the field. Our SEAL approach has been rigorously evaluated on three real-world datasets, demonstrating its superior performance. It consistently outperforms state-of-the-art methods by 7.8% to 22.6% in MCC and 3.9% to 8.4% in Macro-F1. Furthermore, SEAL performance is agnostic to the data encoding framework utilized, enhancing performance by 4.7% to 73.3% in MCC and 2.6% to 30.5% in Macro-F1 across different data encoding models. This robust performance opens up new possibilities for integrating more advanced LMs into CA-HAR tasks. We share our code and supplement material at https://github.com/GMouYes/SEAL.