The development of novel activity recognition approaches requires access to high quality, annotated datasets. When designing an experiment, researchers may not have access to the complete set of equivalent sensors required. A viable solution to this barrier has been the use of completely simulated environments. Nevertheless, an optimal solution may be to allow researchers to use the equipment they do have, and simulate missing sensors. This paper aims to address this scenario through the proposal of the Intelligent Environment Experiment Assistance Tool. The approach facilitates real-time partial simulation of an environment in addition to a real-time annotation component which aims to maximize the accuracy and consistency of dataset an-notations. The concept has received feedback from 19 international re-searchers who are involved in Intelligent Environments research. This feed-back provides an insight into the prevalence, type and impact of limitations in physical IE usage. Additionally, 84.3% of the researchers stated that the real-time annotation would be very useful or quite useful, and 52.6% stated that the partial simulation component would be very useful or quite useful.
- Dataset annotation
- Intelligent environments
Synnott, J., Gabrielli, C., & Nugent, C. (2016). The Intelligent Environment Experiment Assistance Tool to Facilitate Partial Environment Simulation and Real-Time Activity Annotation. In Ubiquitous Computing and Ambient Intelligence (LNCS) (Vol. 10070, pp. 91-97). Springer. https://doi.org/10.1007/978-3-319-48799-1_11