An integrated Mandarin speech database for improving evaluation and rehabilitation procedures of children with hearing impairment

This paper presents the attempts for creating a new Mandarin speech database from which materials for evaluation and training of the hearing impaired children can be selected automatically. The database incorporates computational analyses of corpora, children's language development knowledge, and acoustic features of Mandarin syllables. The selection rules of materials not only meet the phonetic balanced criterion but also take the differences among developmental stages into consideration.