Abstract:
Technology utilized to automatically monitor Activities of Daily Living (ADL) could be a key component in identifying deviations from normal functional profiles and provi...Show MoreMetadata
Abstract:
Technology utilized to automatically monitor Activities of Daily Living (ADL) could be a key component in identifying deviations from normal functional profiles and providing feedback on interventions aimed at improving health. However, if activity recognition systems are to be implemented in real world scenarios such as health and wellness monitoring, the activity sensing modality must unobtrusively fit the human environment rather than forcing humans to adhere to sensor specific conditions. Modern smart phones represent a ubiquitous computing device which has already undergone mainstream adoption. In this paper, we investigate the feasibility of using a modern smartphone, with limited placement constraints, as the sensing modality for an activity recognition system. A dataset of 4 subjects performing 7 activities, using varying sensor placement conditions, is utilized to investigate this. Initial experiments show that a decision tree classifier performs activity classification with precision and recall scores of 0.75 and 0.73 respectively. More importantly, as part of this initial experiment, 3 main problems, and subsequently 3 solutions, relating to unconstrained sensor placement were identified. Using our proposed solutions, classification precision and recall scores were improved by +13% and +14.6% respectively.
Published in: 2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society
Date of Conference: 28 August 2012 - 01 September 2012
Date Added to IEEE Xplore: 10 November 2012
ISBN Information:
ISSN Information:
PubMed ID: 23366641