Visualization of wearable sensor data during swimming for performance analysis

Sensor-based biomechanical monitoring of sporting activity requires the interpretation of large data-sets of time series data-sets. Visualization techniques are a powerful method for displaying these data in a meaningful way to assist in understanding the complex interrelationships of the data and biomechanics. In particular, repetitive actions such as seen in many sports, including swimming can benefit from such analysis where overlay and visual comparison of multiple strokes can be advantageous. Many other disciplines, such as medicine visualize repetitive data and are translational opportunities for the investigation of biomechanical data, such as swimming. This paper presents a case study in which inertial sensor time series data from an elite and sub-elite swimmer were compared using visualization techniques to highlight differences in their action and performance. In particular, the metrics of body roll velocity was captured from the gyroscope sensor and was used as the key time series data to be visualized. Visualization techniques investigated were time-series overlay, phase space portraits, ribbon plot overlay, and wavelet scalograms. The phase space portraits, ribbon plots, and wavelet scalograms demonstrated clearly self-consistency of the swimmer's action. As a cross-comparison tool, these techniques showed clear difference between the elite swimmer, who had lower variability and thus a more consistent action than the sub-elite swimmer. This paper has demonstrated that there is merit in further examination of these techniques as a tool for feedback. It was found that all the methods presented unique views of stroke biomechanics in a nontechnical yet intuitive way for clearer communication.