An Experiment in Hierarchical Recognition of Group Activities Using Wearable Sensors

Abstract

Abstract. Pervasive computing envisions implicit interaction between people and their intelligent environments instead of individual devices, inevitably leading to groups of individuals interacting with the same intelligent environment. These environments must therefore be aware not only of user contexts and activities, but the contexts and activities of groups of users as well. This poster will demonstrate an experiment conducted towards understanding hierarchical multi-user group activity recognition using wearable sensors. The experiment will explore different data abstraction levels in terms of recognition rates, power consumption and wireless communication volumes for the devices involved.

Recognizing Group Activities Using Wearable Sensors

“…Since GAR using wearable sensors is inherently a hierarchical problem, these same issues are also present here as well, but with focus on GAR instead of context prediction. Here a case study on GAR to evaluate the optimal context abstraction level for GAR using sensors from wearable devices is presented, as was described in a preliminary poster abstract [8]. The results provide insight into the power-accuracy trade-off for GAR, and uncover several research questions for the field of GAR in general.…”

A Practical Evaluation of Information Processing and Abstraction Techniques for the Internet of Things

“…An Experiment in Hierarchical Recognition: Gordon et al [23] present an experiment to recognize group activities such as meetings, presentations and coffee breaks. The authors differentiate their work on different levels of abstractions, from lower-level abstractions e.g sensor measurements and mediumlevel abstractions e.g activities such as walking to higher-level abstractions such as meeting or coffee break.…”

“…The authors introduce a system that creates data abstractions from sensor measurements and links it with physical objects and phenomena that are represented in a model. The model is divided into different layers, similar to the work by Gordon et al [23]. Higher layers represent more-abstracted information and include the context of the physical object.…”

“…The ontology is divided into different levels of abstraction, namely lowerlevel and higher-level abstractions. Lower-level abstractions represent single events measured by a particular sensor, higherlevel abstractions are aggregated inferred abstractions incorporating several data sources and processing steps based on the model by [23]. The model uses a clustering algorithm to find similar events to generate the first unlabelled lower-level abstractions.…”