Exploiting harmonic structures to improve separating simultaneous speech in under-determined conditions | IEEE Conference Publication | IEEE Xplore