An active semi-supervised deep learning model for human activity recognition
Haixia Bi,Miquel Perello-Nieto,Raul Santos-Rodriguez,Peter Flach,Ian Craddock
DOI: https://doi.org/10.1007/s12652-022-03768-2
IF: 3.662
2022-03-20
Journal of Ambient Intelligence and Humanized Computing
Abstract:Abstract Human activity recognition (HAR), which aims at inferring the behavioral patterns of people, is a fundamental research problem in digital health and ambient intelligence. The application of machine learning methods in HAR has been investigated vigorously in recent years. However, there are still a number of challenges confronting the task, where one significant barrier lies in the longstanding shortage of annotations. To address this issue, we establish a new paradigm for HAR, which integrates active learning and semi-supervised learning into one framework. The main idea is to reduce the annotation cost by actively selecting the most informative samples for annotation, as well as leveraging the unlabelled instances in a semi-supervised way. In particular, we propose to utilize the massive unlabelled data via temporal ensembling of convolutional neural networks (CNN), which yields robust consensus predictions by aggregating the outputs of the training networks on different epochs. We conducted extensive experiments on three public benchmark datasets. The proposed method achieves Macro F1 values of 0.76, 0.45 and 0.91 in a low annotation scenario on PAMAP2, USCHAD and UCIHAR datasets respectively, outperforming a multitude of state-of-the-art deep models. The ablation study proves the effectiveness of the two components of the framework, i.e., active learning-based sample selection and semi-supervised model training with temporal ensembling, in alleviating the issue of insufficient labels. Cross-validation and statistical significance experiments further demonstrate the robustness and generalization ability of the proposed method. The source codes are available at https://github.com/HaixiaBi1982/ActSemiCNNAct.
computer science, information systems,telecommunications, artificial intelligence