Self-Adaptive and Lightweight Real-Time Sleep Recognition With Smartphone

• DOI: 10.24138/jcomss.v14i3.584
• Publication Date: 2018-09-11T05:04:26Z
• Authors: Ennio Gambi; Simone Barbetta; Adelmo De Santis; Manola Ricciuti
• Keywords: Activity Recognition; Sleep

It is widely recognized that sleep a basic phys- iological process having fundamental effects on human health, performance and well-being. Such evidence stimulates the re- search of solutions to foster self-awareness personal sleeping habits, correct living environment management policies encourage sleep. In this context, use mobile technologies powered with automatic recognition capabilities can be helpful, ubiquitous computing devices like smartphones leveraged as proxies unobtrusively analyse behaviour. To aim, we propose real-time methodology relied smartphone equipped app exploits contextual usage information infer habits. During an initial training stage, selected features are processed by k-Nearest Neighbors, Decision Tree, Random Forest, Support Vector Machine classifiers, select best performing one. Moreover, 1st-order Markov Chain applied improve performance. Experimental results, both offline in Matlab environment, online through fully functional Android app, demonstrate effectiveness proposed approach, achieving acceptable results term Precision, Recall, F1-score.