# Step-Length Estimation in Asymmetric Gait Using a Single Lower-Back IMU Data and a Biomechanical Model Inspired by a Double Inverted Pendulum

## TL;DR

This paper introduces a new model for estimating step length during asymmetric gait using a single sensor on the lower back, showing strong accuracy and potential for clinical use.

## Contribution

The novel biomechanical model incorporates pelvic rotation for more accurate step-length estimation in asymmetric gait.

## Key findings

- The model achieved low Median Absolute Errors (MdAE) below 0.04 m for participants without gait impairment.
- Strong correlation (R = 0.96) and clinically trivial mean bias (0.64 cm) were confirmed with a reference system.

## Abstract

Step length is a fundamental parameter for gait assessment, reflecting complex neuromuscular and biomechanical behavior. Accurate step-length estimation is clinically relevant for monitoring populations with neurological or musculoskeletal conditions, as well as older adults. This study presents a novel biomechanical model, inspired by the inverted double pendulum, for step-length estimation under asymmetric gait conditions using a single inertial sensor on the lower back. Unlike models that assume symmetry, the proposed model explicitly incorporates pelvic rotation, enabling more accurate step length estimation, particularly in individuals with gait impairment. The model was validated against a gold standard OptiTrack® (Corvallis, OR, USA) system with 33 adults: 21 participants without and 12 with gait impairment. Results show that the model achieved low Median Absolute Errors (MdAE), below 0.04 m in participants without gait impairment and remaining within 0.06 m in those with impairment. Statistical validation confirmed a strong correlation with the reference system (R = 0.96, R2 = 0.93) and a clinically trivial mean bias (0.64 cm) from Bland-Altman analysis. These results validate the model’s effectiveness under various gait conditions, suggesting its technical feasibility and strong potential for clinical and real-world applications, particularly for the longitudinal monitoring of patients with functional impairments.