Manifold learning for user profiling and identity verification using motion sensors: gait datasets

The two public gait datasets used to evaluate the proposed method: RecodGait v2 and IDNet's dataset. The user-centric coordinate transformation proposed by our research team in a previous work was applied to these data and the resulting data are in the `user_coordinates` folder. The raw data of RecodGait v2 dataset are in the `raw_data` folder. We are also making available the generated motion-level frames yielded by the pre-processing stage of the proposed method using the experimental setup defined in the paper. These motion-level frames are split into sets, which were used to train and evaluate the proposed method. These motion-level frames are separated by dataset, split into sets (`train_cnn`, `gallery`, `probe`), and there is one file for each axis of the accelerometer (`x`, `y`, `z`).