Multimodal walker-type gait analysis system with encoder data of walker and human pose estimation data using fisheye-camera
| Autor: | Katsuhiko NISHIZAWA, Yuma KOYAMATSU, Toru TSUMUGIWA, Ryuichi YOKOGAWA |
|---|---|
| Jazyk: | japonština |
| Rok vydání: | 2024 |
| Předmět: | |
| Zdroj: | Nihon Kikai Gakkai ronbunshu, Vol 90, Iss 937, Pp 23-00240-23-00240 (2024) |
| Druh dokumentu: | article |
| ISSN: | 2187-9761 67908160 |
| DOI: | 10.1299/transjsme.23-00240 |
| Popis: | The novel posture analysis system for a user's whole body while walking with a walker was developed by using a deep learning method based on the image data of only one camera. The original walker equipped with the system was proposed in this paper. The fisheye-camera enables us to capture the posture of the user's whole body using only one camera equipped on the walker, which is conventionally difficult. (1) The 3-dimensional pose of the user while walking with the walker is estimated from the obtained data using deep learning. (2) The 3-dimensional pose is estimated from the 2-dimensional pose estimation of the user's pose which is obtained from the image data for each of the two cameras. (3) The user's pose is measured using the 3-dimensional position measurement device in order to evaluate the estimation results of (1) and (2). The walking speed and stride-length are estimated using the multilayer perceptron in the multimodal method in which the input data is the estimated results of the user's pose and the moving distances obtained from the encoder equipped in the walker wheel. The estimated results are also evaluated on the measurement results of the 3-dimensional position measurement device. The present system is useful in a practical walker for the rehabilitation of walking, because it is equipped with only one camera, and it enables us to estimate not only the user's posture, but also the user's walking-speed and stride-length during walking with the walker. |
| Databáze: | Directory of Open Access Journals |
| Externí odkaz: |
|