Height-Adaptive Human Gait Model Through Radar-Driven Pipeline With Two Co-Located mmWave MIMO Radars

Roy Veld*, Andre B.J. Kokkeler, Alessandro Chiumento, Yang Miao

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

23 Downloads (Pure)


In this paper, a methodology for developing a human walking model adapted to the individual measured by radar was proposed. The fundamental parameters used in the model are the gait parameters and physical dimensions. Empirical validation of the proposed methodology is undertaken, involving the acquisition of data using a 77-GHz FMCW radar. The data is collected from three distinct individuals walking five different trajectories with respect to the radar. Moreover, the gait parameter estimation accuracy is evaluated for the different walking trajectories of the target. The studied gait parameters are speed, step length, and step frequency. These could be estimated with mean errors up to 0.077 m/s, 9.3 cm, and 0.128 Hz for all trajectories, respectively. Nevertheless, these errors diminish to 0.022 m/s, 2.2 cm and 0.03 Hz, respectively when the targets walk in a straight trajectory aligned with the radar beam. Moreover, the feasibility of estimating body part dimensions directly from the radar data is investigated. It was found that only the total human height could be directly estimated using the employed hardware. Except for the tallest participant of 2.01 m, the height could be estimated with a mean absolute error up to 10.9 cm. Enhanced hardware configurations or the integration of machine learning techniques may improve the accuracy of body part dimension estimations.

Original languageEnglish
Pages (from-to)7199-7223
Number of pages25
JournalIEEE Access
Publication statusPublished - 5 Jan 2024


  • Doppler signatures
  • FMCW radar
  • Human gait
  • Millimeter-wave
  • Walking model


Dive into the research topics of 'Height-Adaptive Human Gait Model Through Radar-Driven Pipeline With Two Co-Located mmWave MIMO Radars'. Together they form a unique fingerprint.

Cite this