Classification of walking speed based on bidirectional LSTM
Walking speed is a powerful predictor of health events which are related to musculoskeletal disorder and mental disease. One of the established computerized technique which employed to perform the gait analysis is motion analysis system. This system allows researchers to perform quantification or es...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Published: |
Springer Science and Business Media Deutschland GmbH
2022
|
| Subjects: | |
| Online Access: | http://eprints.um.edu.my/43458/ https://www.scopus.com/inward/record.uri?eid=2-s2.0-85129251151&doi=10.1007%2f978-3-030-90724-2_7&partnerID=40&md5=3ce36df390017414e97246157cd8229f |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Walking speed is a powerful predictor of health events which are related to musculoskeletal disorder and mental disease. One of the established computerized technique which employed to perform the gait analysis is motion analysis system. This system allows researchers to perform quantification or estimation on human pose and body shape from multiple camera with or without markers. However, it was reported that the high degree of variability within the data representation of gait has resulted important patterns to be undetectable. Through this study, we have developed a stacked bidirectional LSTM (Bi-LSTM) to interpret human walking speed based on kinematic data. A Bi-LSTM has higher training capability compared to a unidirectional LSTM, whereby it enables additional training by traversing the data forward and backward. We employed this model to classify the gait patterns of different walking speeds from 27 sets of gait data with total of 453 gait cycles collected from the walking trial, captured via via Vicon Motion System (Vicon MX, Oxford Metrics, UK). Kinematic parameters of the gait cycles were employed as the input layer of the Bi-LSTM deep learning architecture. Our proposed framework has achieved a prediction accuracy of 77 to classify different speed (slow, normal and fast) conditions. It was also observed that with the prediction accuracy is improved with an increased number of stacked Bi-LSTM layers. © 2022, Springer Nature Switzerland AG. |
|---|