- Machine Studying Strategies for Sensor-based Human Exercise Recognition with Information Heterogeneity — A Assessment(arXiv)
Writer : : Xiaozhou Ye, Kouichi Sakurai, Nirmal Nair, Kevin I-Kai Wang
Summary : Sensor-based Human Exercise Recognition (HAR) is essential in ubiquitous computing, analysing behaviours by means of multi-dimensional observations. Regardless of analysis progress, HAR confronts challenges, significantly in knowledge distribution assumptions. Most research usually assume uniform knowledge distributions throughout datasets, contrasting with the various nature of sensible sensor knowledge in human actions. Addressing knowledge heterogeneity points can enhance efficiency, scale back computational prices, and help in growing personalised, adaptive fashions with much less annotated knowledge. This overview investigates how machine studying addresses knowledge heterogeneity in HAR, by categorizing knowledge heterogeneity varieties, making use of corresponding appropriate machine studying strategies, summarizing obtainable datasets, and discussing future problem
