Mobile phone based training application for kayaking

In recent years human physical activity recognition has been developed for health and training in fitness, sports, or medical area. The actual amount of physical activity needed depends on individual's level of fitness and the goals that have been set. Since a smart phone market has seen a dramatic increase and has been close to our lives, the product of this project is developing an application for detecting human physical activity for the Android platform. In this study, a tri-axial accelerometer placed on the smart phone was used to record the acceleration data for human physical activity classification. For the purpose of managing humans’ daily energy expenditure and time, Charging, Uncarried, Walking, Driving, and Active were chosen as target activities for classification. This application uses K-Nearest Neighbour algorithm with a Decision Tree based approach to classify the activities. The results show that the accuracy of determining Charging state was 100% and the Uncarried state was classified in 90.90% accuracy.