New incremental SVM algorithms for human activity recognition in smart homes

• DOI: 10.1007/s12652-022-03798-w
• Publication Date: 2022-03-25T05:03:08Z
• Authors: Yala Nawal; Mourad Oussalah; Belkacem Fergani; Anthony Fleury
• Keywords: Incremental Learning; Activity Recognition

Abstract Smart homes are equipped with several sensor networks to keep an eye on both residents and their environment, interpret the current situation react immediately. Handling large scale dataset of sensory events real time enable efficient interventions is challenging very difficult. To deal these data flows challenges, traditional streaming classification approaches can be boosted by use incremental learning. In this paper, we presented two new Incremental SVM methods improve performance in context human activity recognition tasks. Two feature extraction elaborated refining dependency focusing last event only have been suggested. On other hand, a clustering based approach similarity suggested boost learning algorithms capitalizing relationship between chunk support vectors previous chunk. We demonstrate through simulations major publicly available sets (Aruba Tulum), feasibility improvements performances achieved our proposed over state-of-the-art. For instance, shown that introduced similarity-based 5 9 times faster than terms training performances. Similarly, Last-state method induces at least 5% improvement F1-score when using baseline classifier.