Authors:
Rui Varandas
1
;
Duarte Folgado
1
and
Hugo Gamboa
2
Affiliations:
1
Associaç ão Fraunhofer Portugal Research, Rua Alfredo Allen 455/461, Porto, Portugal
;
2
Laboratório de Instrumentaç ão, Engenharia Biomédica e Física da Radiaç ão (LIBPhys-UNL), Departamento de Física, Faculdade de Ciências e Tecnologia, FCT, Universidade Nova de Lisboa, 2829-516 Caparica, Portugal
Keyword(s):
Time Series, Anomaly Detection, Human Motion, Unsupervised Learning, Industry.
Abstract:
In industrial contexts, the performed tasks consist of sets of predetermined movements that are continuously repeated. The execution of improper movements and the existence of events that might prejudice the productive system are regarded as anomalies. In this work, it is proposed a framework capable of detecting anomalies in generic repetitive time series, adequate to handle human motion from industrial scenarios. The proposed framework consists of (1) a new unsupervised segmentation algorithm; (2) feature extraction, selection and dimensionality reduction; (3) unsupervised classification based on Density-Based Spatial Clustering Algorithm for applications with Noise. The proposed solution was applied in four different datasets. The yielded results demonstrated that anomaly detection in human motion is possible with an accuracy of 73±19%, specificity of 74 ± 21% and sensitivity of 74 ± 35%, and also that the developed framework is generic and may be applied in general repetitive tim
e series with little adaptation effort for different domains.
(More)