View in EDS

Nonintrusive Integrated Sensing and Noise-Aware Contrastive Learning for Advanced Building Occupancy Analysis.

Saved in:
Bibliographic Details
Title: Nonintrusive Integrated Sensing and Noise-Aware Contrastive Learning for Advanced Building Occupancy Analysis.
Authors: Huang, Kaiyu1 (AUTHOR) khuang29@stevens.edu, Liu, Kaijian2 (AUTHOR) Kaijian.Liu@stevens.edu
Source: Journal of Computing in Civil Engineering. Jan2026, Vol. 40 Issue 1, p1-19. 19p.
Subjects: Signal denoising, Building operation management, Energy consumption, Occupancy rates, Environmental monitoring, Machine learning
Abstract: Accounting for occupancy information in building operations holds significant promise in enabling occupancy-responsive and energy-efficient operations of buildings to better harness their untapped energy efficiency potential. However, existing methods for building occupancy analysis are limited in the following two critical aspects. From the occupancy sensing perspective, limited efforts have explored the use of nonintrusive integrated sensing to address privacy and adherence challenges in occupancy sensing and to capture complementary sources of data for improved sensing performance. From the occupancy analytics perspective, existing methods are limited in addressing noise in data from nonintrusive integrated sensing and the challenges of interclass similarity and intraclass dissimilarity in the data, when predicting occupant presence and occupancy classes. To address these critical limitations, this paper proposes a novel building occupancy analysis framework. The proposed framework (1) employs a proposed nonintrusive integrated occupancy sensing method that effectively integrates multiple nonintrusive gas, volatile organic compound, and environmental sensors to improve the sensing performance—while neither capturing private personal information nor requiring occupant interaction with sensing devices; and (2) utilizes a proposed noise-aware contrastive learning method for denoising raw data collected by the sensing system and addressing the similarity-dissimilarity challenges to better predict occupant presence and occupancy classes. The proposed framework was implemented and deployed in an office space and classroom space for performance evaluation. The experimental results showed that it achieved an F1 measure of 89.14% in the office and 91.32% in the classroom, when predicting the occupant presence and occupancy classes—demonstrating the potential of the proposed framework in supporting nonintrusive and advanced building occupancy analysis. [ABSTRACT FROM AUTHOR]
Copyright of Journal of Computing in Civil Engineering is the property of American Society of Civil Engineers and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:Accounting for occupancy information in building operations holds significant promise in enabling occupancy-responsive and energy-efficient operations of buildings to better harness their untapped energy efficiency potential. However, existing methods for building occupancy analysis are limited in the following two critical aspects. From the occupancy sensing perspective, limited efforts have explored the use of nonintrusive integrated sensing to address privacy and adherence challenges in occupancy sensing and to capture complementary sources of data for improved sensing performance. From the occupancy analytics perspective, existing methods are limited in addressing noise in data from nonintrusive integrated sensing and the challenges of interclass similarity and intraclass dissimilarity in the data, when predicting occupant presence and occupancy classes. To address these critical limitations, this paper proposes a novel building occupancy analysis framework. The proposed framework (1) employs a proposed nonintrusive integrated occupancy sensing method that effectively integrates multiple nonintrusive gas, volatile organic compound, and environmental sensors to improve the sensing performance—while neither capturing private personal information nor requiring occupant interaction with sensing devices; and (2) utilizes a proposed noise-aware contrastive learning method for denoising raw data collected by the sensing system and addressing the similarity-dissimilarity challenges to better predict occupant presence and occupancy classes. The proposed framework was implemented and deployed in an office space and classroom space for performance evaluation. The experimental results showed that it achieved an F1 measure of 89.14% in the office and 91.32% in the classroom, when predicting the occupant presence and occupancy classes—demonstrating the potential of the proposed framework in supporting nonintrusive and advanced building occupancy analysis. [ABSTRACT FROM AUTHOR]
ISSN:08873801
DOI:10.1061/JCCEE5.CPENG-6781