RT Journal Article
T1 Outlier Detection in Wearable Sensor Data for Human Activity Recognition (HAR) Based on DRNNs
A1 Muñoz Organero, Mario
AB Wearable sensors provide a user-friendly and non-intrusivemechanism to extract user-relateddatathat paves the way to the development of personalized applications. Withinthose applications, humanactivityrecognition (HAR) plays an important role in the characterization of the usercontext. Outlierdetectionmethods focus on finding anomalous data samples that are likely to have beengenerated by adifferentmechanism. This paper combines outlier detection and HAR by introducing a novelalgorithmthat is able both todetect information from secondary activities inside the main activity and toextract datasegments of aparticular sub-activity from a different activity. Several machine learningalgorithms havebeen previouslyused in the area of HAR based on the analysis of the time sequences generatedby wearablesensors. Deeprecurrent neural networks (DRNNs) have proven to be optimally adapted to thesequentialcharacteristics ofwearable sensor data in previous studies. A DRNN-based algorithm is proposed inthispaper for outlierdetection in HAR. The results are validated both for intra- and inter-subjectcases and bothfor outlierdetection and sub-activity recognition using two different datasets. A firstdataset comprising4 majoractivities (walking, running, climbing up, and down) from 15 users is used totrain and validatetheproposal. Intra-subject outlier detection is able to detect all major outliersin the walking activity in thisdataset,while inter-subject outlier detection only fails for one participant executingthe activity in a peculiarway.Sub-activity detection has been validated by finding out and extracting walkingsegments present inthe otherthree activities in this dataset. A second dataset using four different users,a different setting anddifferentsensor devices is used to assess the generalization of results.
PB IEEE
SN 2169-3536
YR 2019
FD 2019-06-05
LK https://hdl.handle.net/10016/31798
UL https://hdl.handle.net/10016/31798
LA eng
NO This work was supported by the ‘‘ANALYTICS USING SENSOR DATA FOR FLATCITY’’ Project (MINECO/ ERDF, EU) funded in partby the Spanish Agencia Estatal de Investigación (AEI) under Grant TIN2016-77158-C4-1-R and in part by the European RegionalDevelopment Fund (ERDF).
DS e-Archivo
RD 1 sept. 2024