Towards a Knowledge Base for Activity Recognition of Diverse Users

Indoor Localization for Personalized Ambient Assisted Living of Multiple Users in Multi-Floor Smart Environments

“…As different users, based on their diversity and the associated internal and external factors, can exhibit different forms of navigation and movement patterns during performing different activities in a given environment, that are crucial for personalized indoor localization, it is important to model the diverse ways in which these activities can be performed by different users. To achieve the same, we use the probabilistic reasoning-based mathematical model proposed in [82] that presents multiple equations to model these different ways by which a complex activity may be performed. These equations, as shown in Equations ( 1)-(3), are based on the concept of analyzing a complex activity in terms of atomic activities, context attributes, other atomic activities, other context attributes, start atomic activities, end atomic activities, start context attributes, and end context attributes.…”

“…where: ζ(t): all the different ways in which a complex activity can be performed by different users Θ(t): all the different ways in which a complex activity can be performed where the specific user performing the activity always reaches the end goal Ψ(t): all the different ways of performing a complex activity where the specific user performing the activity never reaches the end goal A t : atomic activities for a complex activity C t : context attributes for a complex activity A ts : all the start atomic activities for a given complex activity C ts : all the start context attributes for a given complex activity A tE : all the end atomic activities for a given complex activity C tE : all the end context attributes for a given complex activity A tδ : all the core atomic activities for a given complex activity C tδ : all the core context attributes for a given complex activity A tI : all the atomic activities for a given complex activity C tI : all the context attributes for a given complex activity η: all the atomic activities for a given complex activity µ: all the context attributes for a given complex activity ρ: all the A tδ for a given complex activity ω: all the C tδ for a given complex activity Equation ( 1) models all the possible different ways by which a complex activity may be performed by different users based on their diversity and internal as well as external factors affecting the activity, which could include environment-based distractions, false starts, delayed completion, and failed attempts leading to missing one or more A ts , C ts , A tE , C tE , A tδ , C tδ , A tI , and C tI [82]. Equation (2) models all the possible scenarios by which a complex activity may be performed in which the user, irrespective of their diversity or the effect of internal and external factors, would always reach the end goal or the desired outcome.…”

“…All these equations [82] were developed by using probabilistic reasoning principles, concepts from the binomial theorem [83], and by reasoning-based analysis of the dynamic weights for each of A ts , C ts , A tE , C tE , A tδ , C tδ , A tI , and C tI , during each complex activity instance [81]. Based on these equations, the work proposed in [82] considers any instance of a complex activity as a designated set that consists of {A ts , C ts , A tE , C tE , A tδ , C tδ , A tI , C tI , η, µ, ρ, ω, ζ(t), Θ(t), Ψ(t)}, where A ts , C ts , A tE , C tE , A tδ , and C tδ are determined based on a weighted approach that considers the weights associated to different A tI and C tI . Thereafter, η, µ, ρ, and ω are computed by applying probabilistic reasoning principles to the sequence of A tI and C tI for the given complex activity [81].…”

“…As can be seen from Table 1, ζ(t): all the different ways by which this complex activity can be performed by different users = 64, Θ(t): all the different ways by which this complex activity can be performed where the specific user performing the activity always reaches the end goal = 4, and Ψ(t): all the different ways of performing this complex activity where the specific user performing the activity never reaches the end goal = 60. Modeling all these instances for different activities in a given environment, as per the methodology discussed in [82], is important towards understanding the specific navigational and movements patterns for determining the indoor location of the specific user as well as other components of the user's indoor position such as building information, floor description, and spatial orientation in terms of whether the user is inside or outside a given indoor spatial region.…”

An Ambient Intelligence-Based Human Behavior Monitoring Framework for Ubiquitous Environments

“…The challenge is therefore to ensure that such fall detection systems can seamlessly function without being dependent on external factors that could affect its operation or performance metrics. Our framework uses concepts of complex activity recognition [30] and two related works [31,32], as well as taking the context-driven approach outlined in Section 3, for the analysis of diverse components of user interactions performed on context parameters to interpret the dynamics of human behavior and their relationships with the contextual and spatial features of an environment to detect any anomalies that could constitute an emergency. The performance, operation, and functionality of such an approach is independent of the effect of any external factors or conditions, such as floor vibrations, WiFi channel state data, and the distance between the user and the ground.…”