Design and implementation of a music composition application using speech recognition

This paper describes speech recognition as a research area, by studying its application to a non trivial problem such as music composition. The design and implementation process of TamTam Listens, a voice-user interface for music composition, is described. This application was developed using open-source tools, like the speech recognition engine PocketSphinx, the acoustic models provided by Voxforge and the Sugar learning platform. Finally, the results of a usability test performed with TamTam Listens are exposed. These results allow drawing conclusions about speech recognition applicability to user interfaces, and offering recomendations for their design and implementation.