Inner-Cycle Phases Can Be Estimated from a Single Inertial Sensor by Long Short-Term Memory Neural Network in...

Area of Science:

• Biomechanics
• Sports Science
• Machine Learning

Background:

• Cross-country roller-ski skating involves complex cyclic movements.
• Accurate measurement of temporal events and inner-cycle parameters is crucial for performance analysis and injury prevention.
• Existing methods may require multiple sensors or laboratory settings.

Purpose of the Study:

• To develop and validate a novel machine learning method for analyzing cross-country roller-ski skating.
• To determine temporal events and inner-cycle parameters using a single inertial measurement unit (IMU).
• To assess the feasibility of field-based, single-IMU locomotion analysis.

Main Methods:

• A long short-term memory (LSTM) neural network was developed to detect ground contact events for poles and skis.
• Eleven athletes performed roller-ski skating at varying intensities and ski conditions.
• Inertial measurement units (IMUs) were placed on the upper back, lower back, and sternum, with force insoles and poles serving as reference systems.

Main Results:

• The IMU on the upper back yielded the best performance, with temporal event detection errors ranging from -1 to 11 ms (SD 64-70 ms).
• Inner-cycle parameters were calculated with mean errors from -11 to 12 ms (SD 66-74 ms).
• The method achieved 95% detection accuracy for pole events and 87% for ski events.

Conclusions:

• The proposed LSTM-based method is a promising tool for assessing temporal events and inner-cycle phases in roller-ski skating.
• A single IMU can effectively estimate spatiotemporal parameters of human locomotion in field settings.
• This technology has potential applications in sports performance monitoring and biomechanical analysis.