"Recognition of nasality and nasal air flow problems in children with cleft palate and/or velopharngeal incompetence"

Supervised by Professor PD Green

Abstract

Cleft palate is a kind of serious birth defect. Occasionally this defect will cause inconvenience for children to communicate with others by generating extra nasal and nasality air flow though the nose. In another word, there is nasality and nasal air flow problem for child, and they need to accept the physical surgery operation according to the different severity level. However the traditional recognition and evaluation of the nasality and nasal sound is invasive assessment approaching with instrument. How to improve the assessment becomes the challenge for clinic department. Recently many methods were introduced to set up valid and reliable measurement of hypernasality. A considerable research is using machine learning method of speech technology. In this project, we consider building up the application of the machine learning speech models that comparing the sequence of the features vector and generatea measure of similar to individual, the Gaussian Mixture model (GMM) to identify the disorder speech of cleft palate children. All the experiments are designed around the disorder speech identification and reliability of classification.