Estimation of Mechanical Power Output Employing Deep Learning on Inertial Measurement Data in Roller Ski Skating

Abstract

The ability to optimize power generation in sports is imperative, both for understanding and balancing training load correctly, and for optimizing competition performance. In this paper, we aim to estimate mechanical power output by employing a time-sequential information-based deep Long Short-Term Memory (LSTM) neural network from multiple inertial measurement units (IMUs). Thirteen athletes conducted roller ski skating trials on a treadmill with varying incline and speed. The acceleration and gyroscope data collected with the IMUs were run through statistical feature processing, before being used by the deep learning model to estimate power output. The model was thereafter used for prediction of power from test data using two approaches. First, a user-dependent case was explored, reaching a power estimation within 3.5% error. Second, a user-independent case was developed, reaching an error of 11.6% for the power estimation. Finally, the LSTM model was compared to two other machine learning models and was found to be superior. In conclusion, the user-dependent model allows for precise estimation of roller skiing power output after training the model on data from each athlete. The user-independent model provides less accurate estimation; however, the accuracy may be sufficient for providing valuable information for recreational skiers.

Using Wearable Sensors to Estimate Mechanical Power Output in Cyclical Sports Other than Cycling—A Review

“…Alternatively, six studies estimated power using all forces acting on the athlete as a single or multibody model (see Table 2 ) [ 10 , 11 , 12 , 24 , 25 , 26 ]. One study estimated mechanical power by considering the athlete and transportation object as a single rigid body (see Table 2 ) [ 13 ].…”

“…Gloersen et al [ 25 ] and Uddin et al [ 26 ] used a single-body model of the athlete to improve the feasibility of mechanical power estimation in cross-country skiing, which was imitated with roller ski skating (see Figure A3 ). To further simplify the approach, both studies only used kinematic data and estimated the resistive forces to estimate mechanical power.…”

“…The propulsive force ( ) was multiplied with the velocity of the CoM to obtain a mechanical power estimation. Uddin et al [ 26 ] performed their experiments on a treadmill, eliminating air drag. Mechanical power was calculated as the sum of power against gravity and rolling resistance.…”

Inner-cycle Phases can be Estimated from a Single Inertial Sensor by Long Short-term Memory Neural Network in Roller Ski Skating

“…Recently, neural networks have been developed to determine cross-country skiing sub-technique in classical style using gyroscope data from the wrist to determine cycles and accelerometer on the chest to perform the classification [6]. For the skating style, multiples IMUs were used to determine mechanical power using a long short-term memory (LSTM) recurrent neural network [7]. Measurements of the head position that could be measured using a differential global navigation system was used to train a neural network classifier to determine the skating sub-technique [8].…”

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

“…Recently, neural networks have been developed to determine a cross-country skiing sub-technique in the classical style using gyroscope data from the wrist to determine cycles and an accelerometer on the chest to perform the classification [ 6 ]. For the skating style, multiple IMUs were used to determine mechanical power using a long short-term memory (LSTM) recurrent neural network [ 7 ]. Measurements of the head position that could be measured using a differential global navigation system were used to train a neural network classifier to determine the skating sub-technique [ 8 ].…”