A Clustering-based Feature Selection for Automatic Labeling in Human Activity Recognition
Bo-Yan Lin, Yu‐Da Lin · 2022 IEEE 4th Global Conference on Life Sciences and Technologies (LifeTech) · 2022
Due to the popularity of mobile devices and wearable sensors, improving the accuracy of human activity recognition has become a research topic. However, to train an actual model based on machine learning in HAR, it is necessary to collect sensor information of a large number of samples and mark each sample as belonging to a specific activity. Therefore, designing an automatic labeling method is important in HAR research. In this study, we applied the feature selection based on fuzzy C-means particle swarm optimization (FS-FCPSO) to improve the accuracy of the automatic labeling results. Through the FS-FCPSO method, six human activities are automatically marked according to the selected important features, including walking, sitting, standing, upstairs, downstairs, and lying down. The results of these six activities are automatically marked by the FS-FCPSO method. Our results show that the FS-FCPSO method is more suitable for automatic marking in HAR than K-means and fuzzy C-means algorithms.