Skip to content

LocoHub Standard - Technical Specification

Core technical specification for parquet file structure and data organization.

For comprehensive reference including biomechanics, see: Biomechanical Standard | Quick Reference

Data Formats

Time-Indexed Data

Original sampling frequency preserved

  • Format: dataset_time.parquet
  • Structure: Continuous time series data
  • Use case: Temporal analysis, event detection

Phase-Indexed Data

Normalized to 150 points per gait cycle

  • Format: dataset_phase.parquet
  • Structure: 150 points per cycle (0-100%)
  • Use case: Cross-subject comparison, averaging

Variable Naming

See Biomechanical Standard - Variable Naming System for complete variable reference.

Subject Naming Convention

All subjects must follow a strict naming convention to ensure global uniqueness and clear population identification.

Format

<DATASET_CODE>_<POPULATION_CODE><SUBJECT_NUMBER>

  • DATASET_CODE: 2-4 character dataset identifier (e.g., UM21, GT23, PROS)
  • POPULATION_CODE: Standardized population identifier (see table below)
  • SUBJECT_NUMBER: 2-3 digit number (01, 02, ..., 999)

Population Codes

Code Population Description
AB Able-bodied Healthy controls with no impairments
TFA Transfemoral Amputee Above-knee amputation
TTA Transtibial Amputee Below-knee amputation
SCI Spinal Cord Injury Complete or incomplete SCI
CVA Stroke Cerebrovascular accident survivors
PD Parkinson's Disease Diagnosed Parkinson's patients
CP Cerebral Palsy Individuals with cerebral palsy
MS Multiple Sclerosis Diagnosed MS patients
OA Osteoarthritis Significant joint osteoarthritis
ELD Elderly Age > 65, otherwise healthy
PED Pediatric Age < 18
IMP Impaired (Other) Other impairments not listed above

Examples

  • UM21_AB01 - University of Michigan 2021, able-bodied subject 01
  • GT23_AB05 - Georgia Tech 2023, able-bodied subject 05
  • PROS_TFA03 - Prosthetics study, transfemoral amputee 03
  • GAIT_CVA12 - Gait study, stroke patient 12
  • AGING_ELD07 - Aging study, elderly subject 07

Why This Matters

  1. Dataset Concatenation: Prevents ID collisions when combining datasets
  2. Population Analysis: Enables filtering by population type across studies
  3. Clinical Context: Preserves important population information
  4. Research Integrity: Maintains clear data provenance

Required Columns

Metadata (required for all datasets): - subject - Subject identifier following naming convention above - subject_metadata - (Optional) Demographics in key:value format (e.g., "age:45,sex:F,height_m:1.68") - task - Biomechanical category (e.g., level_walking, incline_walking) - task_id - Task variant with primary parameter (e.g., incline_10deg, stair_ascent) - task_info - Metadata in key:value format (e.g., "incline_deg:10,speed_m_s:1.2") - step - Step/cycle number

Phase Data (required for phase-indexed): - phase_ipsi - Gait cycle phase (0-100%) aligned to ipsilateral heel strike

Time Data (required for time-indexed): - time_s - Time in seconds

Standard Variables

See Biomechanical Standard - Comprehensive Variable Reference for complete list of standardized variables.

Task Definitions

See Biomechanical Standard - Task Classification System for complete task hierarchy and metadata specifications.

Sign Conventions

See Biomechanical Standard - Coordinate System & Conventions for detailed sign conventions and coordinate system definitions.

Phase Calculation

Phase-indexed data normalization: 1. Detect gait events (heel strike to heel strike) 2. Normalize each cycle to exactly 150 points 3. Calculate phase percentage: phase_percent = (point_index / 149) * 100

Phase Interpretation: - 0% - Heel strike (start of gait cycle) - ~60% - Opposite heel strike (typical) - 100% - Next heel strike (end of cycle)

Missing Data

Handling: - Missing values: NaN (Not a Number) - Invalid measurements: NaN - No synthetic data generation

Quality Flags (optional): - is_reconstructed_<side> - Boolean flag for filled data - Use true for interpolated/reconstructed values

File Examples

Time-indexed: 

subject,subject_metadata,task,task_id,task_info,step,time_s,knee_flexion_angle_ipsi_rad,hip_moment_contra_Nm
UM21_AB01,"age:25,sex:M,height_m:1.75,weight_kg:70",level_walking,level,"speed_m_s:1.2,treadmill:true",0,0.00,0.123,-0.456
UM21_AB01,"age:25,sex:M,height_m:1.75,weight_kg:70",level_walking,level,"speed_m_s:1.2,treadmill:true",0,0.01,0.126,-0.445
UM21_AB01,"age:25,sex:M,height_m:1.75,weight_kg:70",level_walking,level,"speed_m_s:1.2,treadmill:true",1,1.20,0.120,-0.460

Phase-indexed: 

subject,subject_metadata,task,task_id,task_info,step,phase_ipsi,knee_flexion_angle_ipsi_rad,hip_moment_contra_Nm
GT23_AB03,"age:28,sex:F,height_m:1.68,weight_kg:65",incline_walking,incline_10deg,"incline_deg:10,speed_m_s:1.0,treadmill:true",0,0.0,0.123,-0.456
GT23_AB03,"age:28,sex:F,height_m:1.68,weight_kg:65",incline_walking,incline_10deg,"incline_deg:10,speed_m_s:1.0,treadmill:true",0,0.7,0.126,-0.445
GT23_AB03,"age:28,sex:F,height_m:1.68,weight_kg:65",incline_walking,incline_10deg,"incline_deg:10,speed_m_s:1.0,treadmill:true",0,100.0,0.120,-0.460
GT23_AB03,"age:28,sex:F,height_m:1.68,weight_kg:65",incline_walking,incline_10deg,"incline_deg:10,speed_m_s:1.0,treadmill:true",1,0.0,0.125,-0.458

Validation Requirements

Phase Data Validation: - Exactly 150 points per cycle - Phase values: 0.0 to 100.0 - No gaps in phase progression

Variable Validation: - Joint angles: -π to π radians
- Realistic biomechanical ranges - Consistent units across datasets

For detailed implementation examples, see the Python Tutorial and MATLAB Tutorial.