Robust Gait Event Detection Based on the Kinematic Characteristics of a Single Lower Extremity

The observation of gait phases provides essential information for controller design and performance evaluation of lower extremity wearable robots. Specifically, gait events are often defined and detected to distinguish the transition of gait phases. To achieve this, rule-based gait event detection algorithms detect gait events by utilizing the repetitive features in human walking with very few sensors and simple logic. Besides, many of these algorithms define gait events as characteristic features that are detectable from the sensor measurements. However, conventional methods have not fully considered the correlation between the sensor measurement and characteristics of the human motion. Moreover, these methods were only accurate for a limited condition of human motion, for example, walking only or running without sensor noise. Therefore, in this paper, we propose a gait event detection algorithm considering the full kinematic characteristics of the lower extremity under various gait conditions. The proposed algorithm demonstrates robust performance for both walking and running. Besides, to minimize the time delay and the false information in the detected gait events, this paper also proposes a robust signal dithering algorithm that reduces the sensor noise with a limited phase delay. Overall, the performances of the proposed methods are verified through gait experiments with human subjects.