Anomaly Detection in Activities of Daily Living with Linear Drift

Anomalyq detection in Activities of Daily Living (ADL) plays an important role in e-health applications. An abrupt change in the ADL performed by a subject might indicate that she/he needs some help. Another important issue related with e-health applications is the case where the change in ADL under...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Cognitive computation Jg. 12; H. 6; S. 1233 - 1251
Hauptverfasser: Belmonte-Fernández, Óscar, Caballer-Miedes, Antonio, Chinellato, Eris, Montoliu, Raúl, Sansano-Sansano, Emilio, García-Vidal, Rubén
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York Springer US 01.11.2020
Springer Nature B.V
Schlagworte:
Activities of daily living
Algorithms
Anomalies
Artificial Intelligence
Change detection
Computation by Abstract Devices
Computational Biology/Bioinformatics
Computer Science
Control charts
Dementia
Disease
Drift
Fuzzy sets
Hypotheses
Mixtures
Normal distribution
Older people
Quality control
Quality of life
Random variables
Sensors
Statistical process control
Trends
Validity
Activities of Daily Living
Circular normal distribution
Linear drift
Abrupt change
Anomaly detection
ISSN:1866-9956, 1866-9964
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Anomalyq detection in Activities of Daily Living (ADL) plays an important role in e-health applications. An abrupt change in the ADL performed by a subject might indicate that she/he needs some help. Another important issue related with e-health applications is the case where the change in ADL undergoes a linear drift, which occurs in cognitive decline, Alzheimer’s disease or dementia. This work presents a novel method for detecting a linear drift in ADL modelled as circular normal distributions. The method is based on techniques commonly used in Statistical Process Control and, through the selection of a convenient threshold, is able to detect and estimate the change point in time when a linear drift started. Public datasets have been used to assess whether ADL can be modelled by a mixture of circular normal distributions. Exhaustive experimentation was performed on simulated data to assess the validity of the change detection algorithm, the results showing that the difference between the real change point and the estimated change point was 4.9 0 + 3.17 − 1.98 days on average. ADL can be modelled using a mixture of circular normal distributions. A new method to detect anomalies following a linear drift is presented. Exhaustive experiments showed that this method is able to estimate the change point in time for processes following a linear drift.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1866-9956
1866-9964
DOI:10.1007/s12559-020-09740-6