Compared to mainstream sports, rowing lags concerning the adoption of electronic technologies and 'smart software'. By advancing the technologies used inside the rowing boat, crews can improve the efficiency of the training sessions with more real-time data and eliminate the overhead effort of coaches to analyze the captured data. In this study, a method for extracting features from acceleration waveforms is developed and validated. These features are necessary for using complex algorithms on rowing strokes to automate the analysis process. The features represent the rowing technique phases based on the use of the rower`s body parts. This thesis proposes a stroke detection algorithm to separate individual strokes. First, a Kalman filter is used on the waveform to remove noise. Then, the features are extracted from the strokes with timing relative to the stroke length. The features are validated with correlation matrices and outliers are removed from the dataset. Last, an automated analysis algorithm is developed and validated. The algorithm can differentiate strokes with good or bad techniques with an accuracy of 97% in the acquired dataset. Three methods to identify technical errors are tested with promising results. However, to further develop these methods and algorithms, more participation in the training planning is necessary to record specific stroke rates and power efforts. Because of the lack of collaboration with the coach, there were no strokes recorded between steady state and race pace.
© Copyright 2024 Veröffentlicht von Faculteit Industriële Ingenieurswetenschappen. Alle Rechte vorbehalten.