Challenges in gait training methods and application of machine learning

抄録

In Japan's super-aging society, the number of people requiring nursing care is increasing. In the nursing care field, gait analysis and training involve subjective observation and guidance by physical therapists, which requires specialized knowledge and experience. Therefore, there is a need for more effective gait analysis and training that does not require a physical therapist. Many methods have been proposed at the research level to measure information during gait and feed it to the trainer in real-time, but there are issues in the target value setting method. Considering the remarkable progress in machine learning technology in recent years, several studies have been conducted on gait analysis and learning using feature extraction by machine learning. This paper aims to discuss the challenges in conventional gait analysis and training methods and the effectiveness of machine learning techniques. We first review previous studies on gait analysis and training and feedback learning methods in nursing care settings and discusses their challenges. We also introduce previous studies on gait analysis using machine learning. We then explain our proposed method of setting target values using machine learning that considers individual differences and gait feedback training system and discuss its effectiveness when the training target is stumbling.