- Title
- Differences in swimming stroke mechanics and kinematics derived from tri-axial accelerometers during a 200-IM event in South African national swimmers
- Creator
- Musson, Courtney Ruth
- Subject
- Swimming
- Date Issued
- 2020
- Date
- 2020
- Type
- Thesis
- Type
- Masters
- Type
- MA
- Identifier
- http://hdl.handle.net/10948/46337
- Identifier
- vital:39588
- Description
- Context: Swimming is a highly competitive sport, with elite swimmers and coaches constantly looking for ways to improve and challenge themselves to meet new performance goals. The implementation of technology in swimming has proven to be a vital tool in athlete monitoring and in providing coaches with additional information on the swimmer’s performance. Example of this technology is the use of inertial sensory devices such as tri-axial accelerometers. The accelerometers can be used to provide kinematic information with regards to the swimmer’s stroke rate, stroke length and stroke mechanics. In a typical training session, coaches would have to manually time and count their swimmer’s strokes to be able to gain the kinematic information they require. Hence, the use ofinertial sensory technology, such as accelerometers, would provide the necessary information coaches require, allowing them to concentrate on other performance aspects such as theirswimmer’s technique.Aim and objectives: The aim of this study was to determine the kinematic parameters and swimming stroke mechanics that could be derived from tri-axial accelerometers, during a 200-m individual medley (IM) event in South African national level swimmers. Three objectives were set to meet the aim of the study. The first was to identify and differentiate each of the stroking styles using tri-axial accelerometers. The second was to identify and differentiate the kinematic parametersand stroke mechanicsfor all four strokes using tri-axial accelerometers. The third objective was to implement machine learning to automate the identification and interpretation of the accelerometer data. Method:A quantitative, non-experimental descriptive one group post-test only design was used, in which 15 national level swimmers, of which seven male and eight female (mean ±SD: age: 20.9 ± 2.90 years; height: 173.28 ± 10.61 cm; weight: 67.81 ± 8.09 kg; arm span: 178.21 ± 12.15 cm) were tested. Three anthropometric measures were taken (height, weight and arm span) prior to testing, with two tri-axial accelerometers and Polar V800watch and heart rate belt attached to the swimmers left wrist, upper-back and chest, respectively. All swimmerswere required to perform three main swimming sets: 50-m IM, 100-m variation and 200-mIM. Variousdescriptivestatisticsincluding mean, standard deviation and confidence intervals (95%)were used to describe the data. with further inferential statistics including paired t-test, intra-class correlation and Bland Altman analysis wereused to describe the relationship ivbetween the accelerometer and the manually estimated parameters. Additionally, arepeated measures one-way ANOVA (with post-hoc Tukey HSD test) werealso used in an inter-comparison of the stroke parameters between each of the stroking styles. A confusion matrix wasused to measure the classification accuracy of the machine learning model implemented on the accelerometer data.Results:The accelerometers proved successful in identifyingand discerningthe stroke mechanics for each of the four stroking styles, with the use of video footage to validatethe findings. In the stroke kinematic differentiation, theBland Altman analysisresultsshowed an agreement between themanual method and accelerometer-derived estimates, although a discrepancy was evident for several of the kinematic parameters, with a significant difference found with the estimated lap time, average swimming velocity and stroke rate (paired t-test: p <0.001 for all swim sets). The inter-comparison between the stroke parameters per stroking style showed a significant difference with average swimming velocity (repeated one-way ANOVA: F = 1789.37, p <0.001), averages stroke rate (repeated one-way ANOVA: F = 671.70, p <0.001) and average stroke length (repeated one-way ANOVA: F = 346.46, p<0.001) for the population group tested. Furtheranalysis with post-hoc Tukey HSD test showed no significant difference wereevident for the average swimming velocity(Tukey: p > 0.05for all strokes)andbetween freestyle and backstroke for the average stroke rate and stroke length (Tukey:p = 0.0968 andp = 0.997, respectively).Lastly, the machine learning model found a classification accuracy of 96.6% in identifyingand labelling the stroking styles fromthe accelerometer data.Conclusion: It was shown that the tri-axial accelerometers were successful in the identification and differentiation of all the stroking styles, stroke mechanics and kinematics, although a discrepancy was found with the average swimming velocity, stroke rate and lap time estimations. The machine learning model implemented proved the benefits of using artificial intelligence to ease the data process and interpretation by automatically labelling the accelerometer data. Therefore, the use of tri-axial accelerometers as a coaching aid has major potential in the swimming community. However, further research is required to eliminate the time-consuming data processingand to increasetheaccuracy of the accelerometer in the measurement of all the stroke kinematics.
- Format
- x, 152 leaves
- Format
- Publisher
- Nelson Mandela University
- Publisher
- Faculty of Health Sciences
- Language
- English
- Rights
- Nelson Mandela University
- Hits: 519
- Visitors: 572
- Downloads: 96
Collections
NMMU School of Lifestyle Sciences
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | SOURCE1 | Musson, CR 214008088 Dissertation April 2020.pdf | 3 MB | Adobe Acrobat PDF | View Details Download |