This paper presents a speech synthesis method for people with articulation disorders resulting from athetoid cerebral palsy. For people with articulation disorders, there are duration, pitch and spectral problems that cause their speech to be less intelligible and make communication difficult. In order to deal with these problems, this paper describes a Hidden Markov Model (HMM)-based text-To-speech synthesis approach that preserves the voice individuality of those with articulation disorders and aids them in their communication. For the unstable pitch problem, we use the F0 patterns of a physically unimpaired person, with the average F0 being converted to the target F0 in advance. Because the spectrum of people with articulation disorders is often unstable and unclear, we modify generated spectral parameters from the HMM synthesis system by using a physically unimpaired person s spectral model while preserving the individuality of the person with an articulation disorder. Through experimental evaluations, we have confirmed that the proposed method successfully synthesizes intelligible speech while maintaining the target speaker s individuality.
CITATION STYLE
Ueda, R., Aihara, R., Takiguchi, T., & Ariki, Y. (2015). Individuality-preserving spectrum modification for articulation disorders using phone selective synthesis. In SLPAT 2015 - 6th Workshop on Speech and Language Processing for Assistive Technologies, Proceedings (pp. 118–123). Association for Computational Linguistics (ACL). https://doi.org/10.18653/v1/w15-5120
Mendeley helps you to discover research relevant for your work.