University of Twente Student Theses
On the effects of smoothing on machine learning performance in fatigue detection using sensor data
Verschuren, S.P. (2023) On the effects of smoothing on machine learning performance in fatigue detection using sensor data.
|
PDF
6MB |
| Abstract: | Fatigue detection during a running session is important to decrease the risk of injury. Machine learning in combination with wearable sensor technology is a viable option to detect it, and can be used to provide live feedback to the runner. The raw data is often smoothed to decrease the noise, but the effects of this smoothing can be drastically different, depending on the setup of the sensors and data processing. In this paper, we compare the difference in performance of a machine learning algorithm using smoothed data against non-smoothed data in order to examine the influence of different smoothing setups on the detection of fatigue. This experiment used raw acceleration data from 7 sensors placed on the body during a fatiguing treadmill run. On that data, we applied Butterworth low-pass filters with different orders and cutoff frequencies, and for each sensor, the magnitude of the acceleration was calculated. The data was then segmented into strides and four features were extracted per sensor. Then, twelve combinations of sensors were tested by comparing the performance, in terms of accuracy, of a random forest with the smoothed dataset and the non-smoothed dataset using the 5x2 cross validation t-test. We found that the smoothing could increase or decrease the accuracy by 15% when using a single sensor as input data for the random forest. Moreover, the smoothing did not increase or lower the performance in a significant way (alpha = 0.05) when multiple sensors are used, independent of the configuration of the cutoff frequency and order. Additionally, the accuracy increased in general the more sensors are used. |
| Item Type: | Essay (Bachelor) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 31 mathematics, 54 computer science |
| Programme: | Applied Mathematics BSc (56965) |
| Link to this item: | https://purl.utwente.nl/essays/94468 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page