Variation Word Test and Lexical Unit Selection in Malay Corpus Design for Articulation Disorder Screening with Computerized System

摘要

The purpose of this paper is to discuss the overall quality and achievement of the computerized speech early screening system that aim to overcome the problem of Malay articulation disorder. Few experiments had been done for this early prototype of the system which had been tested in few categories from the context variations of the total target words in the corpus design, multiple lexical unit selection and the specific phonetic placed recognition test. The test had been done in a quiet room and been conducted by using real sample patient voice by collaboration with speech therapist from Hospital Sultanah Aminah (HSA) and Sekolah Kebangsaan Pendidikan Khas (SKPK). The result shows that the best lexical unit to be used in Malay artic disorder screening is the phoneme based with right adjustment in FE technique that achieved almost 71.43% of recognition rate an increment for about 4.7% from the baseline result. Overall, 47520 sample voices sample has been collected for the training set in this experiment.