Bodyweight Fitness and Movement Gestures from 50 Participants

9,888 motion-sensor recordings collected in April 2019 by Sven Rosche and Maxim Kuznetsov under the supervision of Philipp Niklas Müller. The data includes 6,045 recordings of fitness exercises from 31 participants and 3,843 recordings of movement gestures from 20 participants, with one participant overlapping both collections. Recording durations range from approximately 1.4 to 3.5 seconds.

Use Cases

• Train models for classifying bodyweight fitness exercises based on motion-sensor recordings.
• Develop algorithms for recognizing simple movement gestures from raw sensor data.
• Benchmark the performance of sensor fusion techniques on human motion data.
• Study the variability of motion patterns across 50 participants with differing recording quality.

Strengths

• 9,888 total recordings provide a substantial sample size for model training.
• Data from 50 anonymized participants offers diversity in movement patterns.
• Clear separation into 'fitness_exercises' (6,045 recordings) and 'movement_gestures' (3,843 recordings) collections.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Recording durations are short, ranging from 1.4 to 3.5 seconds, which may limit analysis of longer movements.
• Data quality varies between recordings, as noted in the metadata files.

Provenance

Source

figshare

Collection Method

Raw motion-sensor recordings collected from participants.

Time Range

April 2019

Freshness

Last updated 2026-05-17 14:35:37; freshness should be verified.