Incoherent and Online Dictionary Learning Algorithm for Motion Prediction

Accurate model development and efficient representations of multivariate trajectories are crucial to understanding the behavioral patterns pedestrian motion. Most existing algorithms use offline learning approaches learn such motion behaviors. However, these cannot take advantage streams data that available after training has concluded, typically not generalizable they have seen before. To solve this problem, paper proposes two for incoherent dictionaries in an online manner by extending augmented semi-non-negative sparse coding (ASNSC) algorithm. We do adding a penalty into objective function promote dictionary incoherence. A trajectory-modeling application is studied, where we consider learned atoms as local primitives. real-world datasets show trained proposed enhanced representation ability converge quickly compared ASNSC. Moreover, well conditioned. In terms trajectory prediction, methods shown be on par (and often better) with state-of-the-art prediction.