Unsupervised Gait Event Identification with a Single Wearable Accelerometer and/or Gyroscope: A Comparison of Methods...

Area of Science:

• Biomechanics
• Wearable Technology
• Sports Science

Background:

• Accurate identification of initial contact (IC) and terminal contact (TC) gait events is crucial for analyzing running mechanics.
• Wearable sensors offer a portable solution for gait analysis, but method selection impacts accuracy.

Purpose of the Study:

• To evaluate 18 methods for identifying IC and TC gait events using single wearable sensors.
• To provide recommendations for optimal method selection based on sensor location (shank or sacrum).

Main Methods:

• Adapted and created code for automatic execution of 18 gait event identification methods.
• Applied methods to data from 74 runners across varied conditions (foot strike, surface, speed).
• Compared estimated gait events against force plate ground truth to quantify error.

Main Results:

• For shank-mounted sensors, Purcell or Fadillioglu methods are recommended for IC, and Purcell for TC.
• For sacrum-mounted sensors, Auvinet or Reenalda methods are recommended for IC, and Auvinet for TC.
• The Lee method achieved 81.9% accuracy for identifying ground contact with sacrum sensors.

Conclusions:

• Specific validated methods exist for accurate running gait event detection using shank or sacrum wearable sensors.
• Method choice significantly influences the precision of initial and terminal contact event identification.
• These findings aid in selecting appropriate algorithms for wearable-based running gait analysis.