Fig 1.
The high-level overview of the interactions in TIHM.
Fig 2.
A sample Markov model with 4 states.
Fig 3.
A framework for detecting Agitation, Irritation and Aggression (AIA) in people with dementia.
Fig 4.
Illustration of the training data.
(a) Demonstration of data collected from an individual’s home over 39 training days. The data is aggregated in 10 minutes intervals and normalised to ensure that the activity level of each sensor is ranged between 0 to 10. The value 0 in the rightmost colour legend corresponding to a dark blue colour indicates no activity and value 10 corresponding to a yellow colour indicates high activity. (b) Aggregated data from 5 sensors; where each 10 minute interval is a five-dimensional array (top). Clustering the aggregated data to map each five-dimensional window to a single state (middle). Dividing the training data into low-active (LA) and high-active (HA) categories (below).
Fig 5.
Illustration of correlation between the total number of detected anomalies and total number of validated notifications for 12 participants by the clinical monitoring team.
Fig 6.
A normal movement pattern reported in the living and hallway for Patient #12.
Fig 7.
Abnormal movement pattern reported in the living and hallway for patient #12 due to presence of guests.
Fig 8.
Receiver operating characteristic curves for multi-occupancy, participant-specific and decision fusion models for the detection of AIA.
Table 1.
Classification insights into decision fusion model.