{"success":true,"data":{"author":"NTNU (Tørring, Refaat, Helbostad, Adde, Ustad); harmonized by Josef Heidler","bucketAccessType":"public","bucketName":"josefheidler/har_children_2024-harth","bucketPrefix":"","category":"Physiology","cloudProvider":"huggingface","createdAt":"2026-11-22T01:09:18.327Z","description":"NTNU-HARChildren, accelerometry for validating human activity recognition (HAR) models in children from NTNU in Trondheim. Children aged 6 to 17, both typically developing (TD) and with cerebral palsy (CP, GMFCS levels I to II), wore two Axivity AX3 triaxial accelerometers (plus or minus 8 g) on the anterior thigh (least-affected side for CP) and the lower back during structured sessions in a laboratory, a gymnasium and outdoors. Thirteen activity types are labelled, including walking, running, jumping, cycling and postures. The data were released at 50 Hz on DataverseNO (DOI 10.18710/EPCXCC); this Hugging Face copy harmonizes them into one Parquet file per participant under harmonized/, covering 78 participants (46 TD, 32 CP files as released), 11.7 million samples and 65 labelled hours. Uses: physical activity measurement in pediatric rehabilitation, wearable sensing for children with motor disability, domain shift between TD and CP gait. Load with Hugging Face datasets or pandas. The DataverseNO record is released under CC0 1.0 (public domain dedication). Cite Tørring et al., Gait & Posture 108:131-137, 2024.","displayTitle":"NTNU-HARChildren: Thigh and Lower-Back Accelerometry of Children with Cerebral Palsy and Typically Developing Children, Activity Recognition","doi":"10.18710/EPCXCC","huggingFaceUrl":"https://huggingface.co/datasets/josefheidler/har_children_2024-harth","id":"700a0fc9-fa18-5119-82fd-ef6d2f8c7416","license":"CC0-1.0","licenseEvidenceUrl":"https://api.datacite.org/dois/10.18710/EPCXCC","objectCountEstimate":80,"probeStatus":"ok","referenceUrl":"https://doi.org/10.1016/j.gaitpost.2023.11.013","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"josefheidler/har_children_2024-harth\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"josefheidler/har_children_2024-harth\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=josefheidler, license:cc0-1.0, size_categories:10M<n<100M, format:parquet, downloads=295"],"format":"mixed","generatedAt":"2026-11-22T01:10:16.056Z","layout":{"example":"harmonized/agile-axolotl.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"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","encoding":"hex","key":"harmonized/agile-axolotl.parquet","truncatedAt":512}],"sampleObjects":["harmonized/agile-axolotl.parquet","harmonized/bold-jacamar.parquet","harmonized/bold-thrush.parquet","harmonized/bold-wombat.parquet","harmonized/brave-baboon.parquet"],"sidecars":[{"bytes":"---\npretty_name: \"Tørring 2024 — thigh + back accelerometry, children (typically developing + cerebral palsy), activity recognition\"\nlicense: cc0-1.0\nlanguage:\n  - en\ntags:\n  - accelerometry\n  - wearable\n  - human-activity-recognition\n  - time-series-classification\n  - thigh-worn\n  - physical-activity\n  - movement-behaviour\n  - time-series\n  - children\n  - cerebral-palsy\ntask_categories:\n  - tabular-classification\n  - other\nsize_categories:\n  - 10M<n<100M\nconfigs:\n  - config_name: default\n    data_files:\n      - split: train\n        path: \"harmonized/*.parquet\"\n---\n\n# Tørring 2024 — thigh + back accelerometry, children (typically developing + cerebral palsy), activity recognition\n\nDual-sensor accelerometry (Axivity AX3, 50 Hz, ±8 g) from the thigh and lower back\nin children with and without cerebral palsy, with activity ground truth for 13 activity\ntypes including walking, running, jumping, cycling, and postural activities. Recordings\nwere collected across lab, gymnasium, and outdoor settings. Harmonized from the Dataverse\nrelease into per-subject parquet — one row per raw accelerometer sample (50 Hz).\n78 participants (46 TD, 32 CP), 11.7 M samples, 100% labelled, 65.0 h labeled.\n\n## Source\n\n- **Paper**: Tørring, B., Refaat, M., Helbostad, J. L., Adde, L., & Ustad, A. (2024).\n  Validation of two novel human activity recognition models for typically developing\n  children and children with cerebral palsy.\n  *Gait & Posture*, *108*, 131–137.\n  https://doi.org/10.1016/j.gaitpost.2023.11.013\n- **Raw data**: https://doi.org/10.18710/EPCXCC (DataverseNO)\n- **License**: CC0 1.0\n\n## Protocol\n\n- **Participants (TD)**: 63 typically developing children in the study (46 in this\n  release — see Harmonization notes). 35 male / 28 female, age 10.5 ± 2.6 years\n  (range 6–15).\n- **Participants (CP)**: 16 in the original paper (32 in this release — see\n  Harmonization notes). 8 male / 8 female, age 11.4 ± 2.2 years (range 8–17),\n  GMFCS levels I–II.\n- **Sensors**: Two Axivity AX3 accelerometers, ±8 g, 50 Hz per participant — anterior midline thigh (midpoint\n  between ASIS and proximal patella; least-affected side for CP participants) and\n  lower back (~L3 vertebra), both mounted with USB connector pointing down.\n- **Sample rate**: 50 Hz (TD recorded at 200 Hz, CP at 100 Hz; the Dataverse release\n  is already downsampled to 50 Hz — no further resampling applied).\n- **Protocol**: Structured sessions conducted in a laboratory, gymnasium, and outdoors.\n  Activities included individual and group sessions with both standardised tasks\n  (walking, running, stair climbing, cycling) and free-play activities (jumping,\n  shuffling, play). Each session covered a predefined set of activities; duration per\n  activity varied by participant.\n\n### Ground-truth provenance\n\nVideo annotation using GoPro Hero 3+ cameras mounted in room corners, with\nframe-by-frame coding in Anvil v6. Four independent raters annotated the recordings;\ninter-rater reliability exceeded 0.95. Sensor–video synchronisation was achieved by\nthe researcher flicking the accelerometer three times at the start of each session,\nproducing a distinct acceleration peak visible in both streams.\n\n## Schema\n\n`timestamp` is epoch-anchored elapsed time: every recording is normalised so its first\nsample falls at `1970-01-01 00:00:00 UTC`. The date is a sentinel — the Axivity AX3\nRTC was not reliably synchronised to wall-clock. Within-participant relative timing is\naccurate at 50 Hz.\n\n| column      | dtype               | notes                                                                 |\n|-------------|---------------------|-----------------------------------------------------------------------|\n| subject     | string              | participant alias (e.g. `agile-axolotl`)                             |\n| timestamp   | datetime64[ns, UTC] | epoch-anchored elapsed time (see note above)                         |\n| thigh_acc_x | float64             | thigh acceleration, corrected frame, g                               |\n| thigh_acc_y | float64             | thigh acce","encoding":"utf8","key":"README.md","name":"README.md","size":10989,"truncated":true}],"sourceObjectKey":"harmonized/agile-axolotl.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"josefheidler/har_children_2024-harth\")\nprint(ds)","starterSamplePath":"harmonized/agile-axolotl.parquet","totalObjects":80},"summary":"NTNU-HARChildren, accelerometry for validating human activity recognition (HAR) models in children from NTNU in Trondheim.","title":"NTNU-HARChildren: Thigh and Lower-Back Accelerometry of Children with Cerebral Palsy and Typically Developing Children, Activity Recognition","updatedAt":"2026-11-22T04:03:41.769Z","visibility":"public","websiteUrl":"https://doi.org/10.18710/EPCXCC"}}