.png){kind=icon}
Return
Towards a Dynamic Inter-Sensor Correlations Learning Framework for Multi-Sensor-Based Wearable Human Activity Recognition
.png){kind=icon}
PRE
AI
DOI：10.1145/3550331.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Multi-sensor-based wearable human activity recognition (WHAR) is a research hotspot in the field of ubiquitous computing. Extracting effective features from multi-sensor data is essential to improve the performance of activity recognition. Despite the excellent achievements of previous works, the challenge remains for modelling the dynamic correlations between sensors. In this paper, we propose a lightweight yet efficient GCN-based dynamic inter-sensor correlations learning framework called DynamicWHAR for automatically learning the dynamic correlations between sensors. DynamicWHAR is mainly composed of two modules: Initial Feature Extraction and Dynamic Information Interaction. Firstly, Initial Feature Extraction module performs data-to-feature transformation to extract the initial features of each sensor. Subsequently, Dynamic Information Interaction module explicitly models the specific interaction intensity between any two sensors, and performs dynamic information aggregation between sensors by the learned interaction intensity. Extensive experiments on four diverse WHAR datasets and two different resource-constrained devices validate that DynamicWHAR outperforms the SOTA models in both recognition performance and computational complexity.
Keywords:
human activity recognition
wearable sensors
graph convolution network
information fusion
Journal
P
IF:
4.5
Papers:
1.1K
Citations:
7.2K
Organization
Cited Papers
Evaluation of the Effect of Systolic Blood Pressure and Pulse Pressure on Cognitive Function: The Women's Health and Aging Study II.png)
PLoS ONE
OAAI
PLoS ONEIF0
OAAIShareSaveMinisterial Advisers and the Politics of Policy‐Making: Bureaucratic Permanence and Popular Control
PREAI
ShareSaveShareSaveShareSave
View more