Applying many-view learning techniques to speech, using different speakers as separate views, to learn more robust acoustic features for speech recognition systems.
Made as a final project for EN.600.479: Representation Learning at Johns Hopkins University.