A framework for occupancy detection and tracking using floor-vibration signals

Drira, Slah; Smith, Ian F. C.

doi:10.1016/j.ymssp.2021.108472

Cited by 12 publications

(7 citation statements)

References 69 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous studies have predominantly utilized the Faster R-CNN method for occupant counting in the realm of computer vision. This method enhances the original R-CNN framework by accelerating performance through shared computation and employing neural networks for region proposal, rather than relying on a selective search [20]. While Faster R-CNN marks an improvement over R-CNN in terms of speed and accuracy, it still falls short in achieving real-time performance, a significant limitation for practical applications [21].…”

Section: Related Workmentioning

confidence: 99%

“…The bottleneck is the same as that in YOLO v5, but the kernel size of the first convolution increases from 1 × 1 to 3 × 3. Based on this data, we can conclude that YOLO v8 is beginning to regress to the ResNet block described in 2015 [20]. The features were concatenated directly into the neck without forcing the same channel dimensions.…”

Section: B Occupant Detectionmentioning

confidence: 99%

“…Occupants in such zones display a diverse range of behaviors, from moving swiftly through the space to remaining stationary for prolonged durations [16] [17][18]. This variability significantly challenges video-based detection systems, which must efficiently track fast-moving individuals and simultaneously accurately count and identify stationary occupants, ensuring comprehensive and precise occupancy detection [19] [20]. Compounding this challenge is the intricate physical layout of under-actuated zones, often marked by obstructions and blind spots arising from furniture, partitions, and varied architectural elements [21].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Occupancy Measurement in Under-Actuated Zones: YOLO-based Deep Learning Approach

Syahputra,

Azhary

et al. 2024

IJACSA

View full text Add to dashboard Cite

The challenge of accurately detecting and identifying individuals within under-actuated zones presents a relevant research problem in occupant detection. This study aims to address the challenge of occupant detection in under-actuated zones through the utilization of the You Only Look Once version 8 (YOLO v8) object detection model. The research methodology involves a comprehensive evaluation of YOLO v8's performance across three distinct zones, where its precision, accuracy, and recall capabilities in identifying occupants are rigorously assessed. The outcomes of this performance evaluation, expressed through quantitative metrics, provide compelling evidence of the efficacy of the YOLO v8 model in the context of occupant detection in under-actuated zones. Across these three diverse under-actuated zones, YOLO v8 consistently exhibits remarkable mean Average Precision (mAP) scores, achieving 99.2% in Zone 1, 78.3% in Zone 2, and 96.2% in Zone 3. These mAP scores serve as a testament to the model's precision, indicating its proficiency in accurately localizing and identifying occupants within each zone. Furthermore, YOLO v8 demonstrates impressive efficiency in executing occupant detection tasks. The model boasts rapid processing times, with all three zones being analyzed in a matter of milliseconds. Specifically, YOLO v8 achieves execution times of 0.004 seconds in both Zone 1 and Zone 3, while Zone 2, which entails slightly more computational effort, still maintains an efficient execution time of 0.024 seconds. This efficiency constitutes a pivotal advantage of YOLO v8, as it ensures expeditious and effective occupant detection in the context of under-actuated zones.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: B Occupant Detectionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Occupancy Measurement in Under-Actuated Zones: YOLO-based Deep Learning Approach

Syahputra,

Azhary

et al. 2024

IJACSA

View full text Add to dashboard Cite

show abstract

“…Specifically, footstep-induced structural vibrations depend both on the person's walking pattern and the shoe type because the shoe serves as the intermediate layer during the force transmission from the foot to the floor. 11,12 Given that it is unrealistic to ask individuals to wear the same footwear every day, this co-dependency leads to difficulty in identifying the owner of the footsteps when multiple people share the same space and each person has multiple pairs of footwear. While existing studies have achieved promising accuracy in person identification using footstep-induced structural vibrations with participants wearing their own shoes, 6,13,14 the effect of shoes on the results has not been explored.…”

Section: Introductionmentioning

confidence: 99%

Robust personalized gait health monitoring through footstep-induced structural vibrations

Dong,

Sun,

Wang

et al. 2024

Sensors and Smart Structures Technologies for Civil, Mechanical, and Aerospace Systems 2024

View full text Add to dashboard Cite

Person identification is important in smart buildings to enable personalized services, such as monitoring individuals' gait health. Existing studies found that the structural vibrations induced by human footsteps provide both identity and gait health information of individuals, such as a person's walking speed, balance, and symmetry, enabling personalized gait health monitoring in smart buildings. However, footstep-induced structural vibrations not only depend on human walking patterns but also on a person's footwear as the footstep force transmits from the foot to the floor. This co-dependency leads to difficulty in identifying the owner of the footsteps when multiple people share the same space and each person has multiple pairs of footwear. In this study, we characterize the effect of footwear on footstep-induced structural vibrations to recognize individuals even when they wear different pairs of shoes (or barefoot). We develop a new metric named Force Transmissibility (FT) that measures the proportion of forces transmitting from the foot to the floor through the footwear. This metric unifies the effect of diverse shoe types, and we utilize this metric to enable robust person identification among various shoe types. We evaluated our approach through real-world walking experiments with eight shoe types shared by four participants. Our method achieves a 22% improvement in identifying the owner of the footsteps when compared to a baseline without footwear considerations.

show abstract

“…Wei deals with the reconstruction of trajectories in urban areas with small sampling frequency [ 23 ]. Drira tracked occupant trajectories using floor-vibration measurements [ 24 , 25 ].…”

Section: Introductionmentioning

confidence: 99%

Calibration of pedestrian ingress model based on CCTV surveillance data using machine learning methods

Pálková,

Uhlík,

Apeltauer

2024

PLoS ONE

View full text Add to dashboard Cite

Machine learning methods and agent-based models enable the optimization of the operation of high-capacity facilities. In this paper, we propose a method for automatically extracting and cleaning pedestrian traffic detector data for subsequent calibration of the ingress pedestrian model. The data was obtained from the waiting room traffic of a vaccination center. Walking speed distribution, the number of stops, the distribution of waiting times, and the locations of waiting points were extracted. Of the 9 machine learning algorithms, the random forest model achieved the highest accuracy in classifying valid data and noise. The proposed microscopic calibration allows for more accurate capacity assessment testing, procedural changes testing, and geometric modifications testing in parts of the facility adjacent to the calibrated parts. The results show that the proposed method achieves state-of-the-art performance on a violent-flows dataset. The proposed method has the potential to significantly improve the accuracy and efficiency of input model predictions and optimize the operation of high-capacity facilities.

show abstract

A framework for occupancy detection and tracking using floor-vibration signals

Cited by 12 publications

References 69 publications

Occupancy Measurement in Under-Actuated Zones: YOLO-based Deep Learning Approach

Occupancy Measurement in Under-Actuated Zones: YOLO-based Deep Learning Approach

Robust personalized gait health monitoring through footstep-induced structural vibrations

Calibration of pedestrian ingress model based on CCTV surveillance data using machine learning methods

Contact Info

Product

Resources

About