GaitVibe Dataset: Footstep-Induced Floor Vibration Data with 3D Motion Capture and Pressure Insole Measurements from 42 Typically Developing Adults

We present GaitVibe, the first footstep-induced floor vibration dataset designed for human gait anaysis, recorded from 42 typically developing adults who walk with their normal gait and/or three simulated abnormal gaits, along with multimodal ground truths of 1) spatio-temporal gait parameters, 2) 3D Motion Capture data (gait kinematics), and 3) pressure insole data (gait kinetics).

Gait analysis is essential for various medical applications, including the diagnosis and rehabilitation of neuromusculoskeletal disorders, assessment of fall risks, and the advancement of sports performance, aiming to guide optimal treatments that enhance an individual's mobility and physical performance. Gait analysis typically involves the measurement of spatio-temporal gait parameters, joint angles, and ground reaction forces, which are all covered in this dataset. Traditional gait analysis is conducted in specialized gait laboratories, requiring a 3D motion capture system and force plates, which produce accurate results, but are very expensive, time-consuming, and require professionally trained medical experts to operate. Due to the limited number of such specialized laboratories, patients often have to travel long distances for these assessments, posing barriers for disadvantaged populations to access quality healthcare. To increase access to gait analysis, various new sensing devices, such as cameras, pressure mats, and inertial measurement units (IMUs), have been developed to conduct gait analysis in daily living environments. However, they are limited by privacy concerns, area of coverage, and discomfort in carrying devices, which is not practical for long-term, continuous monitoring in everyday life.

To overcome the abovementioned limitations, this dataset presents a novel gait analysis approach using ambient floor vibrations. When human footsteps land on the floor, they exert forces onto the floor structures that generate vibrations. These vibrations propagate through the floor and are measured by vibration sensors mounted on the floor surface. This vibration-based approach is low-cost, non-invasive, and contact-free, exhibiting its potential to enable accessible gait analysis in daily living spaces.

The dataset is collected by four sparsely deployed geophone sensors attached to the floor surface along a 7-meter-long walkway, with simultaneous 3D motion capture and pressure insole data as ground truths. To validate the ambient floor vibration data in detecting gait abnormalities, the dataset includes 4 common gait types ``simulated'' by the healthy adult participants under the guidance of biomechanists, including 1) typically developing gait (normal gait), 2) flexed-knee gait, 3) toe-walking gait, and 4) gait with foot drag. This dataset provides a foundation for future advancements in gait monitoring in daily living, which has the potential to lower the barrier to clinical gait analysis and enable long-term tracking of daily gait changes. 

Related publications on this dataset include:

•       Gait parameter estimation: https://pubmed.ncbi.nlm.nih.gov/38676114/
•       Foot-floor contact modeling: https://ascelibrary.org/doi/10.1061/JENMDT.EMENG-7639
•       Gait abnormality detection: https://ieeexplore.ieee.org/document/10556733
•       Joint motion estimation: https://www.sciencedirect.com/science/article/abs/pii/S0022460X25004249?via%3Dihub