Paper ID
1908
Paper Type
short
Description
Gait analysis is a common technique used to identify problems related to movement and posture in people with injuries, and foot-side detection is one of its important challenges. As many commercial sensors only provide limited information and traditional lab-based gait analysis is expensive, the aim of this study is to discriminate between left and right foot steps based on acceleration data from a single chest-worn accelerometer. To achieve this goal, an experimental study was conducted with 25 participants wearing an accelerometer on their chest and walking in a static environment. Several machine learning (ML) classifiers were trained to detect a foot-side from collected acceleration data. All machine learning classifiers achieved high classification accuracy, with Random Forest providing the best results. This result shows that ML-based foot-side classification using a single sensor is achievable and can contribute to develop an efficient health monitoring system to track lower limb’s problems.
Recommended Citation
Choi, Jungyeon; Youn, Jong-hoon; and Haas, Christian, "Machine Learning Approach for Foot-side Classification using a Single Wearable Sensor" (2019). ICIS 2019 Proceedings. 6.
https://aisel.aisnet.org/icis2019/is_health/is_health/6
Machine Learning Approach for Foot-side Classification using a Single Wearable Sensor
Gait analysis is a common technique used to identify problems related to movement and posture in people with injuries, and foot-side detection is one of its important challenges. As many commercial sensors only provide limited information and traditional lab-based gait analysis is expensive, the aim of this study is to discriminate between left and right foot steps based on acceleration data from a single chest-worn accelerometer. To achieve this goal, an experimental study was conducted with 25 participants wearing an accelerometer on their chest and walking in a static environment. Several machine learning (ML) classifiers were trained to detect a foot-side from collected acceleration data. All machine learning classifiers achieved high classification accuracy, with Random Forest providing the best results. This result shows that ML-based foot-side classification using a single sensor is achievable and can contribute to develop an efficient health monitoring system to track lower limb’s problems.