Methods and systems for disentangling discriminative features of a user of a device from motion signals and authenticating a user on a mobile device are provided. In at least one aspect of the methods and systems, each captured motion signal is divided into segments. The segments are then converted into translated segments using one or more trained translation algorithms. The segments and translated segments are then provided to a machine learning system. Discriminative features of the user are then extracted from the segments and translated segments with the processor using the machine learning system that applies one or more feature extraction algorithms.