Humans naturally develop walking capability that is energy efficient, stable, environment adaptive and robust. Lower limb amputations, unfortunately, disrupt this ability; individuals with lower limb amputations usually depend on prosthetic devices to restore the basic walking function. Lower-limb robotic prosthetics can benefit from context awareness to provide enhanced comfort and safety to the amputee. In this work, we aim to develop a terrain identification system based on inertial measurement units IMU streams collected from the lower limb. The system for a prosthetic leg uses visual and inertial sensors though, but we are willing to observe if terrain identification without the visual data is viable.
This is a classification task to find different terrains from time series data. We will use F1 score as the evaluation metric of this project.