Development of iot-based agility timer prototype and classification of agility
To date, there is limited specific device available that can measure agility time and deficient study has been conducted to study agility classification. Thus, the aim of this study is to develop an Internet of Things (IoT)-based agility timer prototype with appropriate agility experiment proto...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English English |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://eprints.uthm.edu.my/6456/1/24p%20NG%20CHUN%20KEAT.pdf http://eprints.uthm.edu.my/6456/2/NG%20CHUN%20KEAT%20COPYRIGHT%20DECLARATION.pdf http://eprints.uthm.edu.my/6456/3/NG%20CHUN%20KEAT%20WATERMARK.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | To date, there is limited specific device available that can measure agility time and
deficient study has been conducted to study agility classification. Thus, the aim of this
study is to develop an Internet of Things (IoT)-based agility timer prototype with
appropriate agility experiment protocol to evaluate the agility time of combat sports
athletes and perform agility profiling using supervised machine learnings. The main
components of the prototype consisted of an Arduino NodeMCU board, a vibration
sensor, an organic light-emitting diode (OLED), three visual stimuli (red, green and
yellow LEDs) and an audio stimulus (buzzer). Through the integration with the Blynk
app, the data obtained can be viewed not only on the OLED display but on Blynk App
too. Prototype assessment by means of statistical analysis was found to be valid (R =
0.998, R
2 = 0.997, p < 0.05), reliable (ICC ≥ 0.9) and accurate (0.06 - 0.084 RMSE).
Fifty combat sports athletes (26 Silat and 24 Taekwondo athletes) were recruited to
undergo two agility experiments: Simple Agility Time (SAT) and Multiple-Choice
Agility Time (MCAT). It was found that 80 % of the participants were more responsive
towards the audio stimulus as compared with the visual stimulus. In terms of visual
cognition, 40 % of the subjects were more responsive towards the red LED stimulus
in comparison with the yellow LED and green LED stimuli. Next, supervised Support
Vector Machine (SVM), K-Nearest Neighbor (KNN) and Artificial Neural Network
(ANN) were implemented to classify agility time into three classes, which were high,
medium and low based on two inputs: agility time and body mass index (BMI). The
classification benchmark was determined based on the agility time threshold range.
The findings revealed that the best supervised classifier model was ANN, which gave
100 % accuracy for each stimulus. Next, an agility calculator based on the ANN model
was developed to obtain the athletes’ agility class. In conclusion, a valid, reliable and
accurate IoT-based agility timer prototype was successfully developed to assess the
agility time of combat sports athletes, and an agility calculator based on the ANN
model was created to obtain the agility class of athletes. |
|---|