Footstep-Induced Structural Vibration Data for Person Identification with 8 Walking Speeds

Footstep-induced structural vibration data from 10 participants walking at 8 different speeds, collected in 2017 and 2021. The dataset contains vertical velocity signals from 5 geophone sensors at a 1000Hz sampling rate, stored as footstep traces in a MAT-file. It was created by researchers including Yiwen Dong and Shijia Pan from Stanford University and is hosted on Zenodo.

Use Cases

• Training person identification models based on footstep-induced vibration patterns.
• Studying the impact of walking speed variation on structural vibration signatures.
• Developing sensor fusion algorithms using data from multiple geophone sensors.
• Benchmarking time-series classification methods for biometric applications.

Strengths

• Data from 10 participants across 8 controlled walking speeds, providing variability.
• Signals captured by 5 geophone sensors at a 1000Hz sampling rate.
• Three layers of labels: person identity, sensor number, and walking speed.

Limitations

• Row count and exact dataset size are unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect bias inherent to the specific 30ft x 6ft concrete hallway and participant demographics (8 male, 2 female, aged 20-29).

Provenance

Source

Zenodo, via paperswithcode. Original research by Yiwen Dong, Shijia Pan, et al.

Collection Method

Experimental data collected from human participants walking in a controlled hallway environment, with structural vibrations sensed by geophones.

Time Range

Data associated with research from 2017 and a 2021 dataset release.

Geography

Experiment conducted in a specific indoor location (hallway); geographic scope is not specified.