Using Sedentary Activity Classification Model to Illustrate an Adaptable Freelance Workspace*

Freelancing and remote-work is fast gaining acceptance in the global community, and thus resulting into sedentary behaviours that poses threat to healthy living. Recent studies have shown that several health hazards are related to sedentary behaviors. Specifically, too much sitting is adversely associated with health outcomes, of which includes cardio-metabolic risk biomarkers, type 2 diabetes and premature mortality. In this paper a one dimensional(1D) convolutional neural network based sedentary activity classification model is proposed for classification of sit, stand and sit-stand transitions using sensor data fusion collected with a tri-axial Inertial Measurement Unit (IMU) sensor. Specifically, normalized data from the IMU axes within a specific time window were considered to classify sedentary behaviors (sit, stand and sit-stand transition). The precision of the classification of the three studied sedentary behaviors exceeded 96.8% using IMU data fusion. The applicability of the proposed classification model is extended to a simulated condition of a typical freelancer by modeling an adaptable automated workspace.