A Framework for the Identification of Human Vertical Displacement Activity Based on Multi-Sensor Data - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue IEEE Sensors Journal Année : 2022

A Framework for the Identification of Human Vertical Displacement Activity Based on Multi-Sensor Data

Ajaykumar Manivannan
  • Fonction : Auteur
Elias Willemse
  • Fonction : Auteur
Balamurali T.
  • Fonction : Auteur
Wei Chien Benny Chin
  • Fonction : Auteur
Yuren Zhou
  • Fonction : Auteur
Bige Tuncer
  • Fonction : Auteur
Roland Bouffanais
  • Fonction : Auteur
  • PersonId : 1130023

Résumé

With new cities increasingly expanding vertically, there is a pressing need to shed light on human vertical mobility, which can readily be achieved with existing sensor technology. To date, the methodology to track and identify vertical movement from large-scale unstructured data sets is lacking. Here, we design and develop such a framework to accurately and systematically identify the sparse human vertical displacement activity that is typically buried into the predominantly horizontal mobility. Our framework uses sensor data from barometer, accelerometer and Wi-Fi scanner coupled with an extraction step involving a combination of feature engineering and data segmentation. This methodology is subsequently integrated into a machine-learning-based classifier to automatically distinguish vertical displacement activity from its horizontal counterpart. We confirm the high accuracy of this approach by a thorough validation and testing showing a 98% overall accuracy and a 92% F1-score in classifying vertical displacement activity.We illustrate the potential of the developed framework by applying it to an unstructured large-scale data set associated with over 16,000 participants going about their daily activity in the city-state of Singapore. This gives us access to all the vertical movements of this large population, and we investigate the statistical distribution of vertical activity, both in terms of number of events and size of vertical jumps, and their temporal heterogeneity across the day. The approach developed here could be used in massive human experiments to uncover the hidden patterns of human vertical mobility. This new knowledge would have significant ramifications for the architectural design of vertical cities.
Fichier principal
Vignette du fichier
V2_manuscript___Multi_sensor_Identification_of_Vertical_Displacement_Activity.pdf (6.55 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03604682 , version 1 (10-03-2022)

Identifiants

Citer

Ajaykumar Manivannan, Elias Willemse, Balamurali T., Wei Chien Benny Chin, Yuren Zhou, et al.. A Framework for the Identification of Human Vertical Displacement Activity Based on Multi-Sensor Data. IEEE Sensors Journal, In press, 22 (8), pp.8011. ⟨10.1109/JSEN.2022.3157806⟩. ⟨hal-03604682⟩
38 Consultations
72 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More