Anomaly Detection in Smart Houses for Healthcare

• Published in: SN computer science 2024-01, Vol.5 (1), p.136
• Main Authors: Galvão, Yves M, Castro, Letícia, Ferreira, Janderson, Neto, Fernando Buarque de Lima, Fagundes, Roberta Andrade de Araújo, Fernandes, Bruno J. T
• Format: Article
• Language: English
• Subjects: Activities of daily living; Anomalies; Context; Datasets; Deep learning; Health care; Health risk assessment; Health risks; Houses; Keywords; Machine learning; Smart houses; Surveillance; Systematic review; Traffic surveillance; Workplaces
• ISSN: 2662-995X; 2661-8907
• DOI: 10.1007/s42979-023-02480-y

Nowadays, device monitoring is an activity present in various different environments. Ranging from monitoring workers in their workplaces, city traffic, surveillance in shops, to elderly at home, all that rely on effective anomaly detection in video scenes. In the context of residences, although there are many kinds of monitoring cameras and sensors, these devices usually are not able to detect health risks automatically. The traditional methods of monitoring people in a house to avoid potential health risks are expensive and, in most cases, require the healthcare professional’s physical presence. A possible alternative for this problem is using a machine learning model to detect health risks by monitoring daily activities. Although these models are capable of identifying activities that represent health risks, many of them depend on labeled data to identify and classify such events. Moreover as these events rarely occur, the sought models have to be effective to avoid needing labeled data. This paper presents a systematic review of the anomaly detection models in smart houses related to identifying health risks. A special attention was given to anomaly detection approaches that avoid using labeled data. After applying the proposed protocol in five databases, between 2009 and 2023 (June), we have identified 1185 studies have met the quality criteria. The selected papers were analyzed using an ad hoc questionnaire, and were ranked according to their relevance. The results suggest that anomaly detection is an important research area in the context of smart houses related to health risks, and bring some insights into why it is expanding in the recent years.