Consonant enhancement for articulation disorders based on non-negative matrix factorization

We present consonant enhancement on a voice for a person with articulation disorders resulting from athetoid cerebral palsy. The movement of such speakers is limited by their athetoid symptoms, and their consonants are often unstable or unclear, which makes it difficult for them to communicate. Speech recognition for articulation disorders has been studied; however, its recognition rate is still lower than that of physically unimpaired persons. In this paper, an exemplar-based spectral conversion using Non-negative Matrix Factorization (NMF) is applied to consonant enhancement of a voice with articulation disorders. The source speaker's spectrum is easily converted into a well-ordered speaker's spectrum. Its effectiveness is examined for voice quality and clarity of consonants for a person with articulation disorders.