An Exploration of Machine Learning Methods for Gait Analysis of Potential Guide Dogs

Wu Y, Nichols C, Foster M, et al. An Exploration of Machine Learning Methods for Gait Analysis of Potential Guide Dogs ACI '23: Proceedings of the Tenth International Conference on Animal-Computer Interaction. 2024;10:1-15. https://doi.org/10.1145/3637882.3637883.

This study investigates the use of machine learning models to analyze gait patterns in potential guide dogs using inertial measurement data from wearable sensors. Two experiments were conducted, examining the effects of sensor placement on gait analysis accuracy, with models achieving classification accuracies ranging from 42% to 91%. Error rates for gait parameter prediction varied from 14% to 29%, with neck-mounted sensors providing more movement data than back-mounted ones. Despite a drop in performance during cross-dataset testing, the models showed potential in generalizing gait-specific behaviors across different dogs.