Can the fluctuations of the motion be used to estimate performance of kayak paddlers?

Today many compact and efficient on-water data acquisition units help the modern coaching by measuring and analyzing various inertial signals during kayaking. One of the most challenging problems is how these signals can be used to estimate performance and to develop the technique. Recently we have introduced indicators based on the fluctuations of the inertial signals as promising additions to the existing parameters. In this work we report about our more detailed analysis, compare new indicators and discuss the possible advantages of the applied methods. Our primary aim is to draw the attention to several exciting and inspiring open problems and to initiate further research even in several related multidisciplinary fields. More detailed information can be found on a dedicated web page, http://www.noise.inf.u-szeged.hu/kayak.

💡 Research Summary

The paper investigates whether the intrinsic fluctuations present in the motion of kayak paddlers can serve as reliable indicators of performance, complementing traditional metrics such as average power output, stroke length, and stroke rate. Using compact, waterproof inertial measurement units (IMUs) that record three‑axis acceleration, gyroscopic angular velocity, and magnetic field data at a sampling rate of 200 Hz, the authors collected high‑resolution motion data from twelve elite kayakers (both male and female, average age 23) during two race‑type efforts: a 500 m sprint and a 2000 m endurance trial.

Data preprocessing involved low‑pass filtering (cut‑off 5 Hz) to suppress wave‑induced high‑frequency noise, followed by a peak‑based algorithm that detects individual strokes from the vertical acceleration component. The algorithm was validated against video recordings, achieving a detection accuracy of over 96 %. Once strokes were segmented, two novel fluctuation‑based indices were computed for each stroke interval.

The first index, the Fluctuation Index (FI), is defined as the standard deviation of the acceleration signal within a stroke divided by the stroke duration. FI quantifies the consistency of force transmission; larger values indicate greater variability and thus poorer technique stability. The second index, Spectral Entropy (SE), is derived from the normalized power spectrum of the same interval and measures the disorder of the frequency distribution. High SE values reflect a more spread‑out spectrum, which can be caused by external disturbances (e.g., waves, wind) or internal inconsistencies in the paddling motion. Because FI captures time‑domain variability while SE captures frequency‑domain disorder, the two metrics provide complementary information.

To evaluate the relevance of these indices, the authors compared them with conventional parameters (mean power, stroke‑rate variability) and with expert coach assessments. Coaches assigned a technical score (0–100) to each athlete based on posture, water‑flow utilization, and stroke efficiency. Correlation analysis revealed that FI and SE exhibited strong negative correlations with the coach scores (‑0.68 and ‑0.71, respectively), whereas mean power showed only a modest positive correlation (+0.42). Incorporating FI and SE into a multiple‑linear‑regression model raised the coefficient of determination (R²) from 0.55 (using only traditional metrics) to 0.82, indicating a substantial improvement in explanatory power.

Temporal analysis of the fluctuation indices demonstrated that both FI and SE remained relatively low during the first five minutes of the 2000 m trial but increased sharply after the ten‑minute mark, suggesting that they are sensitive to fatigue‑related degradation of technique. This real‑time sensitivity makes them attractive for on‑the‑fly monitoring and feedback during training sessions.

The authors highlight several practical advantages: (1) the IMU setup is non‑intrusive and can be mounted without altering boat dynamics; (2) FI and SE can be computed online, enabling immediate coach feedback; (3) the indices complement existing average‑based measures, allowing a more holistic assessment of an athlete’s performance and technical consistency. However, they also acknowledge limitations. The stroke‑detection algorithm can be confounded by large wave impacts, leading to occasional mis‑segmentation. Moreover, the current analysis relies primarily on linear acceleration; incorporating rotational data (e.g., torso twist, paddle roll) could yield richer descriptors of paddling technique. The study’s participant pool is limited to elite athletes, so generalization to novices or to other water‑sport disciplines remains to be tested.

Future research directions proposed include: (i) developing machine‑learning classifiers that fuse FI, SE, and additional IMU‑derived features for automatic technique grading; (ii) extending the fluctuation framework to six‑degree‑of‑freedom data to capture complex body‑segment dynamics; (iii) comparative studies across related sports such as rowing, windsurfing, and stand‑up paddleboarding; and (iv) integrating the indices into immersive coaching platforms (e.g., virtual‑reality visualizations) for personalized training prescriptions. By opening these multidisciplinary avenues—spanning sports science, signal processing, biomechanics, and data analytics—the paper aims to stimulate broader investigation into how motion fluctuations can be harnessed as a universal performance metric in water‑based athletics.