Automated and robust stride segmentation using an instrumented walking cane

Abstract

Human gait analysis is an important tool for assessing mobility and identifying mobility-related disorders; however, clinical gait assessment is often performed in specialized laboratories, limiting accessibility and real-world applicability. This work introduces a robust approach to stride segmentation for an instrumented walking cane, combining gyroscope and strain-gauge signals. A ground-truth dataset was collected using an instrumented cane and motion capture in a Computer-Assisted Rehabilitation Environment laboratory. Data from twelve able-bodied participants included 6,207 annotated strides across five inclines. A novel Extended-Pattern subsequence Dynamic Time Warping (XP-sDTW) algorithm was developed, employing a longer, symmetric step pattern compared to the conventional version. Using a leave-one-subject-out evaluation framework, XP-sDTW achieved an F-score of 95.8 ± 4.5%. The results demonstrate that a single, pre-trained template can effectively segment cane-assisted gait across users and slopes. Future work will extend validation to real-world environments and clinical populations, enabling accessible gait monitoring and rehabilitation.