Self-Supervised Representation Learning With MUlti-Segmental Informational Coding (MUSIC)