A Fuzzy Spatio-temporal-Based Approach for Activity Recognition

Abstract

Abstract. Over the last decade, there has been a significant deployment of systems dedicated to surveillance. These systems make use of real-time sensors that generate continuous streams of data. Despite their success in many cases, the increased number of sensors leads to a cognitive overload for the operator in charge of their analysis. However, the context and the application requires an ability to react in real-time. The research presented in this paper introduces a spatio-temporal-based approach the objective of which is to provide a qualitative interpretation of the behavior of an entity (e.g., a human or vehicle). The process is formally supported by a fuzzy logic-based approach, and designed in order to be as generic as possible.

Fuzzy human motion analysis: A review

“…In the literature of activity recognition, there exists some works [147,148] that employed the FIS to classify different human activities.…”

“…Both [147,148] took into account the uncertainties in both the spatial and temporal features for efficient human behavior recognition. Their method aims at handling high uncertainty levels and the complexities occurring in the real world applications.…”

“…Their method aims at handling high uncertainty levels and the complexities occurring in the real world applications. [147] used the spatial and temporal geometry features to study the importance of the spatiotemporal relations such as 'IsMoving', 'IsComingCloseTo', 'IsGoingAway', 'IsGoingAlong' with the objective to provide a qualitative interpretation of the behavior of an entity (e.g. a human) in real-time.…”

“…[147,148] FIS effectively distinguishes the human motion patterns and activity recognition with its flexibility in customizing the membership functions and the fuzzy rules with tolerance to the vague feature data.…”

A Big Bang–Big Crunch Type-2 Fuzzy Logic System for Machine-Vision-Based Event Detection and Summarization in Real-World Ambient-Assisted Living

“…However, such wearable devices are intrusive and could be uncomfortable and inconvenient as the deployment of wearable devices is invasive for the skin and muscles of the users. T1FLS was used in [20], [21] to analyse the spatial and temporal features for efficient human behaviour recognition. However, there are intra-and inter-subject variations in behavioural characteristics which cause high levels of uncertainty.…”