Publications
Detailed Information
Gait Pattern Prediction and Gait-Based Human Identification Using Machine Learning Methods : 기계학습 기반 기법을 이용한 보행 동작 예측과 보행 기반 개인 식별
Cited 0 time in
Web of Science
Cited 0 time in Scopus
- Authors
- Advisor
- 박종우
- Major
- 공과대학 기계항공공학부
- Issue Date
- 2018-08
- Publisher
- 서울대학교 대학원
- Description
- 학위논문 (박사)-- 서울대학교 대학원 : 공과대학 기계항공공학부, 2018. 8. 박종우.
- Abstract
- This thesis is concerned with human gait motion learning and prediction,
focusing on differences in gait motion between individuals to address the
problems of personalized gait pattern prediction and gait based biometric
identification.
We first present an algorithm that predicts individual gait patterns from
a human's anthropometric characteristics, e.g., age, height, weight, gender,
and other related anthropometric data. We adopt the Gaussian process
dynamical model~(GPDM) framework to address the high dimensionality
of human gait motions, and to learn a common stochastic dynamics model
for gaits. We also utilize Gaussian process regression~(GPR) to learn
a mapping from the space of body features to the motion parameters in
the GPDM framework. Using our framework, an entire cycle of individualized
gait motions at arbitrary walking speeds can be predicted from body feature
data.
We also propose a gait-based identification algorithm based on deep neural
networks. To allow for new subjects whose information is not included in the
original database, we develop a novel autoencoder architecture that is
designed to extract features efficiently from the measured gait motions.
We use the reconstruction error of the trained autoencoder as a similarity
measure between the learned motion and the input, for which user-specified thresholds can
be used for the identification. Our proposed method is able to identify both
subjects in the database as well as unknown subjects, only using the observed
gait motion. Furthermore, it is possible to update the database incrementally
for newly given gait motions.
Finally, we propose an algorithm used to predict a subject's age, height, weight, and
gender from the subject's gait motion. We construct a neural network architecture
for this purpose by modifying the previous autoencoder structure used for the identification, and
devise an efficient learning strategy utilizing transfer learning to reduce the
prediction error.
To validate proposed algorithms, we collect gait motions and anthropometric data
from more than a hundred subjects. For our gait motion prediction algorithm, an individualized
gait pattern is generated for observed anthropometric data and at arbitrary walking speeds.
The algorithm also shows 30% less gait pattern prediction errors compared to a frame-by-frame
statistical regression method. In the case of gait identification, our algorithm
shows identification accuracy of 98.5-100%, outperforming the results reported in recent studies.
- Language
- English
- Files in This Item:
Item View & Download Count
Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.